Next: No warranty, Previous: (dir), Up: (dir) [Contents][Index]
This manual (23 June 2022) is for Libmarpa 9.0.3.
Copyright © 2022 Jeffrey Kegler.
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Next: About this document, Previous: Top, Up: Top [Contents][Index]
The Libmarpa license takes precedence over the statements in this document. In particular, the license states that Libmarpa is free software and has no warranty. No statement in this document should be construed as providing any kind of warranty.
Next: About Libmarpa, Previous: No warranty, Up: Top [Contents][Index]
• How to read this document: | ||
• Prerequisites: | ||
• Parsing theory: | ||
• Terminology and notation: |
Next: Prerequisites, Previous: About this document, Up: About this document [Contents][Index]
This is essentially a reference document, but its early chapters lay out concepts essential to the others. Readers will usually want to read the chapters up and including Introduction to the method descriptions in order. Otherwise, they should follow their interests.
Next: Parsing theory, Previous: How to read this document, Up: About this document [Contents][Index]
This document is very far from self-contained. It assumes the following:
Marpa::R2
or Marpa::R3
,
in Perl.
Next: Terminology and notation, Previous: Prerequisites, Up: About this document [Contents][Index]
This document assumes some acquaintance with parsing theory. The reader’s level of knowledge is probably adequate if he can answer the following questions, either immediately or after a little reflection.
Previous: Parsing theory, Up: About this document [Contents][Index]
In this document,
max(x,y)
is the maximum of
x
and y
,
where x
and y
are two numbers.
• Application and diagnostic behavior: |
Previous: Terminology and notation, Up: Terminology and notation [Contents][Index]
An application behavior is a behavior on which it is intended that the design of applications will be based. Most of the behaviors specified in this document are application behaviors. We sometimes say that “applications may expect” a certain behavior to emphasize that that behavior is an application behavior.
After an irrecoverable failure, the behavior of a Libmarpa application is undefined, so that there are no behaviors which can be relied on for normal application processing, and therefore, there are no application behaviors. In this circumstance, some of the application behaviors become diagnostic behaviors. A diagnostic behavior is a behavior which it is suggested that the programmer may attempt in the face of an irrecoverable failure, for testing, diagnostics and debugging. They are hoped for, rather than expected, and intended to allow the programmer to deal with irrecoverable failures as smoothly as possible. (See Failure.)
In this document, a behavior is a diagnostic behavior only if that is specifically indicated. Applications should not be designed to rely on diagnostics behaviors. We sometimes say that “diagnostics may attempt” a certain behavior to emphasize that that behavior is a diagnostic behavior.
Next: Architecture, Previous: About this document, Up: Top [Contents][Index]
Libmarpa implements the Marpa parsing algorithm. Marpa is named after the legendary 11th century Tibetan translator, Marpa Lotsawa. In creating Marpa, I depended heavily on previous work by Jay Earley, Joop Leo, John Aycock and Nigel Horspool.
Libmarpa implements the entire Marpa algorithm. This library does the necessary grammar preprocessing, recognizes the input, and produces parse trees. It also supports the ordering, iteration and evaluation of the parse trees.
Libmarpa is very low-level. For example, it has no strings. Rules, symbols, and token values are all represented by integers. This, of course, will not suffice for many applications. Users will very often want names for the symbols, non-integer values for tokens, or both. Typically, applications will use arrays to translate Libmarpa’s integer ID’s to strings or other values as required.
Libmarpa also does not implement most of the semantics. Libmarpa does have an evaluator (called a “valuator”), but it does not manipulate the stack directly. Instead, Libmarpa, based on its traversal of the parse tree, passes optimized step by step stack manipulation instructions to the upper layer. These instructions indicate the token or rule involved, and the proper location for the true token value or the result of the rule evaluation. For rule evaluations, the instructions include the stack location of the arguments.
Marpa requires most semantics to be implemented in the application. This allows the application total flexibility. It also puts the application is in a much better position to prevent errors, to catch errors at runtime or, failing all else, to successfully debug the logic.
Next: Input, Previous: About Libmarpa, Up: Top [Contents][Index]
• Major objects: | ||
• Time objects: | ||
• Reference counting: | ||
• Numbered objects: |
Next: Time objects, Previous: Architecture, Up: Architecture [Contents][Index]
The classes of Libmarpa’s object system fall into two types: major and numbered. These are the Libmarpa’s major classes, in sequence.
The major objects have one letter abbreviations, which are used frequently. These are, in the standard sequence,
Next: Reference counting, Previous: Major objects, Up: Architecture [Contents][Index]
All of Libmarpa’s major classes, except the configuration class, are “time” classes. Except for objects in the grammar class, all time objects are created from another time object. Each time object is created from a time object of the class before it in the sequence. A recognizer cannot be created without a precomputed grammar; a bocage cannot be created without a recognizer; and so on.
When one time object is used to create a second time object, the first time object is the parent object and the second time object is the child object. For example, when a bocage is created from a recognizer, the recognizer is the parent object, and the bocage is the child object.
Grammars have no parent object. Every other time object has exactly one parent object. Value objects have no child objects. All other time objects can have any number of children, from zero up to a number determined by memory or some other machine-determined limit.
Every time object has a base grammar. A grammar object is its own base grammar. The base grammar of a recognizer is the grammar that it was created with. The base grammar of any other time object is the base grammar of its parent object. For example, the base grammar of a bocage is the base grammar of the recognizer that it was created with.
Next: Numbered objects, Previous: Time objects, Up: Architecture [Contents][Index]
Every object in a “time” class has its own, distinct, lifetime, which is controlled by the object’s reference count. Reference counting follows the usual practice. Contexts which take a share of the “ownership” of an object increase the reference count by 1. When a context relinquishes its share of the ownership of an object, it decreases the reference count by 1.
Each class of time object has a “ref” and an “unref”
method, to be used by those contexts which need to
explicitly increment and decrement the reference count.
For example, the “ref” method for the grammar class is
marpa_g_ref()
and the “unref” method for the grammar class is
marpa_g_unref()
.
Time objects do not have explicit destructors. When the reference count of a time object reaches 0, that time object is destroyed.
Much of the necessary reference counting is performed automatically. The context calling the constructor of a time object does not need to explicitly increase the reference count, because Libmarpa time objects are always created with a reference count of 1.
Child objects “own” their parents, and when a child object is successfully created, the reference count of its parent object is automatically incremented to reflect this. When a child object is destroyed, it automatically decrements the reference count of its parent.
In a typical application, a calling context needs only to remember to “unref” each time object that it creates, once it is finished with that time object. All other reference decrements and increments are taken care of automatically. The typical application never needs to explicitly call one of the “ref” methods.
More complex applications may find it convenient to have one or more contexts share ownership of objects created in another context. These more complex situations are the only cases in which the “ref” methods will be needed.
Previous: Reference counting, Up: Architecture [Contents][Index]
In addition to its major, “time” objects, Libmarpa also has numbered objects. Numbered objects do not have lifetimes of their own. Every numbered object belongs to a time object, and is destroyed with it. Rules and symbols are numbered objects. Tokens values are another class of numbered objects.
Next: Exhaustion, Previous: Architecture, Up: Top [Contents][Index]
• Earlemes: | ||
• The basic models of input: | ||
• Terminals: |
Next: The basic models of input, Previous: Input, Up: Input [Contents][Index]
• The traditional input model: | ||
• The latest earleme: | ||
• The current earleme: | ||
• The furthest earleme: |
Next: The latest earleme, Previous: Earlemes, Up: Earlemes [Contents][Index]
In traditional Earley parsers, the concept of location is very simple. Locations are numbered from 0 to n, where n is the length of the input. Every location has an Earley set, and vice versa. Location 0 is the start location. Every location after the start location has exactly one input token associated with it.
Some applications do not fit this traditional input model — natural language processing requires ambiguous tokens, for example. Libmarpa allows a wide variety of alternative input models.
In Libmarpa a location is called a earleme. The number of an Earley set is the ID of the Earley set, or its ordinal. In the traditional model, the ordinal of an Earley set and its earleme are always exactly the same, but in Libmarpa’s advanced input models the ordinal of an Earley set can be different from its location (earleme).
The important earleme values are the latest earleme. the current earleme, and the furthest earleme. Latest, current and furthest earleme, when they have determinate values, obey a lexical order in this sense: The latest earleme is always at or before the current earleme, and the current earleme is always at or before the furthest earleme.
Next: The current earleme, Previous: The traditional input model, Up: Earlemes [Contents][Index]
The
latest Earley set
is the Earley set completed most recently.
This is initially the Earley set at location 0.
The latest Earley set is always the Earley set with the highest ordinal,
and the Earley set with the highest earleme location.
The
latest earleme is the earleme of the latest Earley set.
If there is an Earley set at the current earleme,
it is the latest Earley set and the latest earleme
is equal to the current earleme.
There is never an Earley set after the current earleme,
and therefore the latest Earley set is never after the
current earleme.
The marpa_r_start input()
and
marpa_r_earleme_complete()
methods
are only ones that change the latest earleme.
See marpa_r_start_input() and
marpa_r_earleme_complete().
The latest earleme is different from the current earleme if and only if there is no Earley set at the current earleme. A different end of parsing can be specified, but by default, parsing is of the input in the range from earleme 0 to the latest earleme.
Next: The furthest earleme, Previous: The latest earleme, Up: Earlemes [Contents][Index]
The
current earleme
is the earleme that Libmarpa is currently working on.
More specifically, it is the one at which new tokens will start.
Since tokens are never zero length, a new token will always end after the
current earleme.
marpa_r_start_input()
initializes the current earleme to 0,
and every call to
marpa_r_earleme_complete()
advances the
current earleme by 1.
The marpa_r_start input()
and
marpa_r_earleme_complete()
methods
are only ones that change the current earleme.
See marpa_r_start_input() and
marpa_r_earleme_complete().
Previous: The current earleme, Up: Earlemes [Contents][Index]
Loosely speaking,
the
furthest earleme
is the furthest earleme reached by the parse.
More precisely,
it is the highest numbered
earleme at which a token ends
and is 0 if there are no tokens.
The furthest earleme is 0 when a recognizer is
created.
With every call to
marpa_r_alternative()
, the end of the token
it adds is calculated.
A token ends at the earleme location current+length,
where current is the current earleme,
and length is the length of the newly added token.
If old_f
is the furthest earleme before
a call to
marpa_r_alternative()
,
the furthest earleme after the call
is max(old_f, current+length)
.
The marpa_r_new()
and
marpa_r_alternative()
methods
are only ones that change the furthest earleme.
See marpa_r_new() and
marpa_r_alternative().
In the basic input models,
where every token has length 1,
calling
marpa_r_earleme_complete()
after each
marpa_r_alternative()
call is sufficient to process
all inputs,
and the furthest earleme’s value
can be typically be ignored.
In alternative input models,
where tokens have lengths greater than 1,
calling marpa_r_earleme_complete()
once after the last token
is read may not be enough to ensure that all tokens have been processed.
To ensure that all tokens have been processed,
an application must advance the current earleme
by calling
marpa_r_earleme_complete()
,
until the current earleme is equal to the furthest earleme.
For the purposes of presentation, we (somewhat arbitrarily) divide Libmarpa’s input models into two groups: basic and advanced. In the basic input models of input, every token is exactly one earleme long. This implies that, in a basic model of input,
In the advanced models of input, tokens may have a length other than 1. Most applications use the basic input models. The details of the advanced models of input are presented in a later chapter. See Advanced input models.
• The standard model of input: | ||
• Ambiguous input: |
Next: Ambiguous input, Previous: The basic models of input, Up: The basic models of input [Contents][Index]
In the standard model of input,
there is exactly one successful
marpa_r_alternative()
call
immediately previous
to every
marpa_r_earleme_complete()
call.
A marpa_r_alternative()
call is
immediately previous to a
marpa_r_earleme_complete()
call
iff
that marpa_r_earleme_complete()
call is
the first
marpa_r_earleme_complete()
call after
the marpa_r_alternative()
call.
Recall that, since the standard model is
a basic model,
the token length in every successful call to marpa_r_alternative()
will be one.
For an input of length n, there will be
exactly n marpa_r_earleme_complete()
calls,
and all but the last call
to marpa_r_earleme_complete()
must be successful.
In the standard model,
after a successful call
to
marpa_r_alternative()
,
if c is the value of the current earleme before the call,
In the standard model,
a call to
marpa_r_earleme_complete()
follows a successful call of
marpa_r_alternative()
,
so that the value of the furthest earleme before the call to
marpa_r_earleme_complete()
will be c+1
,
where c is the value of the current earleme.
After a successful call to
marpa_r_earleme_complete()
,
c+1
; and
Recall that, in the basic models of input, the latest earleme is always equal to the current earleme.
Previous: The standard model of input, Up: The basic models of input [Contents][Index]
We can loosen the standard model to
allow more than one successful call to
marpa_r_alternative()
immediately previous to each call to
marpa_r_earleme_complete()
.
This change will mean that multiple tokens become possible
at each earleme —
in other words, that the input becomes ambiguous.
We continue to require that there be
at least one successful call to
marpa_r_alternative()
before each call to
marpa_r_earleme_complete()
.
And we recall that,
since this is a basic input model,
all tokens must have a length of 1.
In the ambiguous input model, the behavior of the current, latest and furthest earlemes are exactly as described for the standard model. See The standard model of input.
Previous: The basic models of input, Up: Input [Contents][Index]
A terminal symbol is a symbol which may appear in the input. Traditionally, all LHS symbols, as well as the start symbol, must be non-terminals. This is Marpa’s behavior, by default.
Marpa allows the user to eliminate the distinction between terminals and non-terminals. In this, it differs from traditional parsers. Libmarpa can arrange for a terminal to appear on the LHS of one or more rules, or for a terminal to be the start symbol. However, since terminals can never be zero length, it is a logical contradiction for a nulling symbol to also be a terminal and Marpa does not allow it.
Token values are int
’s.
Libmarpa does nothing with token values except accept
them from the application and return them during
parse evaluation.
A parse is exhausted when it cannot accept any further input. A parse is active iff it is not exhausted. For a parse to be exhausted, the furthest earleme and the current earleme must be equal. However, the converse is not always the case: if more tokens can be read at the current earleme, then it is possible for the furthest earleme and the current earleme to be equal in an active parse.
Parse exhaustion always has a location.
That is, if a parse is exhausted it is exhausted at some earleme location X
.
If a parse is exhausted at location X
, then
X
.
X
.
X
.
X
.
X
.
X
.
X
.
marpa_r_alternative()
after a parser has become exhausted.
X
.
marpa_r_earleme_complete()
after a parser has become exhausted.
Users sometimes assume that parse exhaustion means parse failure. But other users sometimes assume that parse exhaustion means parse success. For many grammars, there are strong associations between parse exhaustion and parse success, but the strong association can go either way, Both exhaustion-loving and exhaustion-hating grammars are very common in practical application.
In an
exhaustion-hating
application,
parse exhaustion typically means parse failure.
C programs, Perl scripts and most programming languages
are exhaustion-hating applications.
If a C program is well-formed,
it is always possible to read more input.
The same is true of a Perl program that does not have a __DATA__
section.
In an exhaustion-loving application parse exhaustion means parse success. A toy example of an exhaustion-loving application is the language consisting of balanced parentheses. When the parentheses come into perfect balance the parse is exhausted, because any further input would unbalance the brackets. And the parse succeeds when the parentheses come into perfect balance. Exhaustion means success. Any language which balances start and end indicators will tend to be exhaustion-loving. HTML and XML, with their start and end tags, can be seen as exhaustion-loving languages.
One common form of exhaustion-loving parsing occurs in lexers which look for longest matches. Exhaustion will indicate that the longest match has been found.
It is possible for a language to be
exhaustion-loving at some points
and exhaustion-hating at others.
We mentioned Perl’s __DATA__
as a complication in a
basically exhaustion-hating language.
marpa_r_earleme_complete()
and
marpa_r_start_input
are the only methods
that may encounter parse exhaustion.
See marpa_r_earleme_complete() and
marpa_r_start_input().
When the marpa_r_start_input
or
marpa_r_earleme_complete()
methods
exhaust the parse,
they generate a MARPA_EVENT_EXHAUSTED
event.
Applications
can also query
parse exhaustion status directly
with the
marpa_r_is_exhausted()
method.
See marpa_r_is_exhausted().
Next: Threads, Previous: Exhaustion, Up: Top [Contents][Index]
Libmarpa handling of semantics is unusual. Most semantics are left up to the application, but Libmarpa guides them. Specifically, the application is expected to maintain the evaluation stack. Libmarpa’s valuator provides instructions on how to handle the stack. Libmarpa’s stack handling instructions are called “steps”. For example, a Libmarpa step might tell the application that the value of a token needs to go into a certain stack position. Or a Libmarpa step might tell the application that a rule is to be evaluated. For rule evalution, Libmarpa will tell the application where the operands are to be found, and where the result must go.
The detailed discussion of Libmarpa’s handling of semantics is in the reference chapters of this document, under the appropriate methods and classes. The most extensive discussion of the semantics is in the section that deals with the methods of the value time class (Value methods).
Libmarpa is thread-safe, given circumstances as described below. The Libmarpa methods are not reentrant.
Libmarpa is C89-compliant. It uses no global data, and calls only the routines that are defined in the C89 standard and that can be made thread-safe. In most modern implementations, the default C89 implementation is thread-safe to the extent possible. But the C89 standard does not require thread-safety, and even most modern environments allow the user to turn thread safety off. To be thread-safe, Libmarpa must be compiled and linked in an environment that provides thread-safety.
While Libmarpa can be used safely across multiple threads, a Libmarpa grammar cannot be. Further, a Libmarpa time object can only be used safely in the same thread as its base grammar. This is because all time objects with the same base grammar share data from that base grammar.
To work around this limitation, the same grammar definition can be used to a create a new Libmarpa grammar time object in each thread. If there is sufficient interest, future versions of Libmarpa could allow thread-safe cloning of grammars and other time objects.
Next: Introduction to the method descriptions, Previous: Threads, Up: Top [Contents][Index]
As a reminder, no language in this chapter (or, for that matter, in this document) should be read as providing, or suggesting the existence of, a warranty. See license. Also, see No warranty.
Next: User non-conformity to specified behavior, Previous: Failure, Up: Failure [Contents][Index]
Libmarpa is a C language library, and inherits the traditional C language approach to avoiding and handling user programming errors. This approach will strike readers unfamiliar with this tradition as putting an appallingly large portion of the burden of avoiding application programmer error on the application programmer themself.
But in the early 1970’s, when the C language first stabilized, the alternative, and the consensus choice for its target applications was assembly language. In that context, C was radical in its willingness to incur a price in efficiency in order to protect the programmer from themself. C was considered to take a excessively "hand holding" approach which very much flew in the face of consensus.
The decades have made a large difference in the trade-offs, and the consensus about the degree to which even a low-level language should protect the user has changed. It seems inevitable that C will be replaced as the low-level language of choice, by a language which places fewer burdens on the programmer, and more on the machine. The question seems to be not whether C will be dethroned as the “go to” language for low-level progamming, but when, and by which alternative.
Modern hardware makes many simple checks essentially cost-free, and Libmarpa’s efforts to protect the application programmer go well beyond what would have been considered best practice in the past. But it remains a C language library. But, on the whole, the Libmarpa application programmer must be prepared to exercise the high degree of carefulness traditionally required by its C language environment. Libmarpa places the burden of avoiding irrecoverable failures, and of handling recoverable failures, largely on the application programmer.
Next: Classifying failure, Previous: Libmarpa's approach to failure, Up: Failure [Contents][Index]
This document specifies many behaviors for Libmarpa application programs to follow, such as the nature of the arguments to each method. The C language environment specifies many more behaviors, such as proper memory management. When a non-conformity to specified behavior is unintentional and problematic, it is frequently called a “bug”. Even the most carefully programmed Libmarpa application may sometimes contain a “bug”. In addition, some specified behaviors are explicitly stated as characterizing a primary branch of the processing, rather than made mandatory for all successful processing. Non-conformity to non-mandatory behaviors can be efficiently recoverable, and is often intentional.
This chapter describes how non-conformity to specified behavior by a Libmarpa application is handled by Libmarpa. Non-conformity to specified behavior by a Libmarpa application is also called, for the purposes of this document, a Libmarpa application programming failure. In contexts where no ambiguity arises, Libmarpa application programming failure will usually be abbreviated to failure.
Libmarpa application programming success in a context is defined as the absence of unrecovered failure in that context. When no ambiguity arises, Libmarpa application programming success is almost always abbreviated to success. For example, the success of an application means the application ran without any irrecoverable failures, and that it recovered from all the recoverable failures that were detected.
Next: Memory allocation failure, Previous: User non-conformity to specified behavior, Up: Failure [Contents][Index]
A Libmarpa application programming failure, unless specified otherwise, is an irrecoverable failure. Once an irrecoverable failure has occurred, the further behavior of the program is undefined. Nonetheless, we specify, and Libmarpa attempts, diagnostics behaviors (see Application and diagnostic behavior) in an effort to handle irrecoverable failures as smoothly as possible.
A Libmarpa application programming failure is recoverable if and only if it is specified as such.
A failure is called a hard failure is it has an error code associated with it. A recoverable failure is called a soft failure if it has no associated error code. (For more on error codes, see Error codes.)
All failures fall into one of five types. In order of severity, these are
Next: Undetected failure, Previous: Classifying failure, Up: Failure [Contents][Index]
Failure to allocate memory is the most irrecoverable of irrecoverable
errors.
Even effective error handling assumes the ability to allocate memory,
so that the practice has been, in the event of a memory allocation failure,
to take Draconian action.
On
memory allocation failure,
as with all irrecoverable failures,
Libmarpa’s behavior in undefined,
but Libmarpa attempts to terminate the current program abnormally by calling abort()
.
Memory allocation failure is the only case in which Libmarpa terminates the program. In all other cases, Libmarpa leaves the decision to terminate the program, whether normally or abnormally, up to the application programmer.
Memory allocation failure does not have an error code. As a pedantic matter, memory allocation failure is neither a hard or a soft failure.
Next: Irrecoverable hard failure, Previous: Memory allocation failure, Up: Failure [Contents][Index]
An undetected failure is a failure that the Libmarpa library does not detect. Many failures are impossible or impractical for a C library to detect. Two examples of failure that the Libmarpa methods do not detect are writes outside the bounds of allocated memory, and use of memory after it has been freed. C is not strongly typed, and arguments of Libmarpa routines undergo only a few simple tests, tests which are inadequate to detect many of the potential problems.
By undetected failure we emphasize that we mean failures undetected by the Libmarpa methods. In the examples just given, there exist tools that can help the programmer detect memory errors and other tools exist to check the sanity of method arguments.
This document points out some of the potentially undetected problems, when doing so seems more helpful than tedious. But any attempt to list all the undetected problems would be too large and unwieldy to be useful.
Undetected failure is always irrecoverable. An undetected failure is neither a hard or a soft failure.
Next: Partially recoverable hard failure, Previous: Undetected failure, Up: Failure [Contents][Index]
An irrecoverable hard failure is an irrecoverable Libmarpa application programming failure that has an error code associated with it. Libmarpa attempts to behave as predictably as possible in the face of a hard failure, but once an irrecoverable failure occurs, the behavior of a Libmarpa application is undefined.
In the event of an irrecoverable failure, there are no application behaviors. The diagnostic behavior for a hard failure is as described for the method which detects the hard failure. At a minimum, this diagnostic behavior will be returning from the method which detects the hard failure with the return value specified for hard failure, and setting the error code as specified for hard failure.
Next: Library-recoverable hard failure, Previous: Irrecoverable hard failure, Up: Failure [Contents][Index]
A partially recoverable hard failure is a recoverable Libmarpa application programming failure
For every partially recoverable hard failure, this document specifies the application behaviors that remain available after it occurs. The most common kind of partially recoverable hard failure is a library-recoverable hard failure. For an example of partially recoverable hard failure, see Library-recoverable hard failure.
Next: Fully recoverable hard failure, Previous: Partially recoverable hard failure, Up: Failure [Contents][Index]
A library-recoverable hard failure is a type of partially recoverable hard failure. Loosely described, it is a hard failure which allows the programmer to continue to use many of the Libmarpa methods in the library, but which disallows certain methods on some objects.
To state the restrictions of application behaviors more precisely, let the “failure grammar” be the base grammar of the method which detected the library-recoverable hard failure. After a library-recoverable hard failure, the following behaviors are no longer applcation behaviors:
Recall that any use of a behavior which is not an application behavior is an irrecoverable failure.
The application behaviors remaining after a library-recoverable hard failure are the following:
An example of a library-recoverable hard failure is
the MARPA_ERR_COUNTED_NULLABLE
error
in the marpa_g_precompute
method.
See marpa_g_precompute().
Next: Soft failure, Previous: Library-recoverable hard failure, Up: Failure [Contents][Index]
A fully recoverable hard failure is a recoverable Libmarpa application programming failure
One example of a fully recoverable hard failure is
the error code MARPA_ERR_UNEXPECTED_TOKEN_ID
.
The “Ruby Slippers” parsing technique
(see Ruby Slippers),
which has seen extensive usage,
is based
on Libmarpa’s ability to recover from
a MARPA_ERR_UNEXPECTED_TOKEN_ID
error
fully and efficiently,
Next: Error codes, Previous: Fully recoverable hard failure, Up: Failure [Contents][Index]
An soft failure is an recoverable Libmarpa application programming failure that has no error code associated with it. Hard errors are assigned error codes in order to tell them apart. Error codes are not necessary or useful for soft errors, because there is at most one type of soft failure per Libmarpa method.
Soft failures are so called, because they are the least severe kind of failure. The most severe failures are “bugs” — unintended, and a symptom of a problem. Soft failures, on the other hand, are a frequent occurrence in normal, successful, processing. In the phrase “soft failure”, the word “failure” is used in the same sense that its cognate “fail” is used when we say that a loop terminates when it “fails” its loop condition. That ”failure” is of a condition necessary to continue on a main branch of processing, and a signal to proceed on another branch.
It is expected that Libmarpa applications will be designed such that successful execution is based on the handling specified for soft failures. In fact, a non-trival Libmarpa application can hardly be designed except on that basis.
Previous: Soft failure, Up: Failure [Contents][Index]
As stated, every hard failure has an associated error code. Full descriptions of the error codes that are returned by the external methods are given in their own section (External error codes).
How the error code is accessed depends on the method which detects the hard failure associated with that error code. Methods for time objects always set the error code in the base grammar, from which it may be accessed using the error methods described below (Error methods). If a method has no base grammar, the way in which the error code for the hard failures that it detects can be accessed will be stated in the description of that method.
Since the error of a time object is set in the base grammar, it follows that every object with the same base grammar has the same error code. Objects with different base grammars may have different error codes.
While error codes are properties of a base grammar, irrecoverability is application-wide. That is, whenever any irrecoverable failure occurs, the entire application is irrecoverable. Once an application becomes irrecoverable, those Libmarpa objects with error codes for recoverable errors are still subject to the general irrecoverability.
Next: Static methods, Previous: Failure, Up: Top [Contents][Index]
The following chapters describe Libmarpa’s methods in detail.
• About the overviews: | ||
• Naming conventions: | ||
• Return values: | ||
• How to read the method descriptions: |
Next: Naming conventions, Previous: Introduction to the method descriptions, Up: Introduction to the method descriptions [Contents][Index]
The method descriptions are grouped into chapters and sections. Each such group of methods descriptions begins, optionally, with an overview. These overviews, again optionally, end with a “cheat sheet”. The “cheat sheets” name the most important Libmarpa methods in that chapter or section, in the order in which they are typically used, and very briefly describe their purpose.
The overviews sometimes speak of an “archetypal” application. The archetypal Libmarpa application implements a complete logic flow, starting with the creation of a grammar, and proceeding all the way to the return of the final result from a value object. In the archetypal Libmarpa application, the grammar, input and semantics are all small but non-trivial.
Next: Return values, Previous: About the overviews, Up: Introduction to the method descriptions [Contents][Index]
Methods in Libmarpa follow a strict naming convention.
All methods have a name beginning with
marpa_
,
if they are part of the
external interface.
If an external method is not a static method,
its name is prefixed with one of
marpa_c_
,
marpa_g_
,
marpa_r_
,
marpa_b_
,
marpa_o_
,
marpa_t_
or
marpa_v_
,
where the single letter between underscores
is one of the Libmarpa major class abbreviations.
The letter indicates which class
the method belongs to.
Methods that are exported,
but that are part of
the internal interface,
begin with _marpa_
.
Methods that are part of the internal interface
(often called “internal methods”)
are subject to change and are intended for use
only by Libmarpa’s developers.
Libmarpa reserves the
marpa_
and _marpa_
prefixes for itself,
with all their capitalization variants.
All Libmarpa names visible outside the package
will begin with a capitalization variant
of one of these two prefixes.
Next: How to read the method descriptions, Previous: Naming conventions, Up: Introduction to the method descriptions [Contents][Index]
Some general conventions for return values are worth mentioning:
NULL
usually indicates failure.
Any other result usually indicates success.
The Libmarpa programmer should not overly rely on the general
conventions for return values.
In particular, -2 may sometimes be ambiguous —
both a valid return value
for success, and a potential indication of hard failure.
In this case, the programmer must distinguish the two return statuses
based on the error code,
and a programmer who is relying too heavily on the general
conventions will fall into a trap.
For a the description of the return values of
marpa_g_rule_rank_set()
,
see Rank methods.
Previous: Return values, Up: Introduction to the method descriptions [Contents][Index]
The method descriptions are written on the assumption that the reader has the following in mind while reading them:
void
,
the last paragraph of its method description is a
“return value summary”.
The return value summary
starts with the label “Return Value”.
Next: Configuration methods, Previous: Introduction to the method descriptions, Up: Top [Contents][Index]
[Accessor] Checks that the Marpa library in use is compatible with the
given version. Generally, the application programmer will pass in the constants
MARPA_MAJOR_VERSION
,
MARPA_MINOR_VERSION
, and
MARPA_MICRO_VERSION
as the three arguments,
to check that their application was compiled with headers
the match the version of Libmarpa that they
are using.
If required_major.required_minor.required_micro is an exact match with 9.0.3, the method succeeds. Otherwise the return status is an irrecoverable hard failure.
Return value: On success, MARPA_ERR_NONE
.
On hard failure, the error code.
[Accessor] Writes the version number in version.
It is an undetected irrecoverable hard failure
if version does not have room for three int
’s.
Return value: Always succeeds. The return value is indeterminate.
Next: Grammar methods, Previous: Static methods, Up: Top [Contents][Index]
The configuration object is intended for future extensions.
These may
allow the application to override Libmarpa’s memory allocation
and fatal error handling without resorting to global
variables, and therefore in a thread-safe way.
Currently, the only function of the Marpa_Config
class is to give
marpa_g_new()
a place to put its error code.
Marpa_Config
is Libmarpa’s only “major”
class which is not a time class.
There is no constructor or destructor, although
Marpa_Config
objects do need to be initialized
before use.
Aside from its own accessor,
Marpa_Config
objects are only used by
marpa_g_new()
and no reference to their location is not kept
in any of Libmarpa’s time objects.
The intent is to that it be convenient
to have them in memory that might be deallocated
soon after
marpa_g_new()
returns.
For example, they could be put on the stack.
[Mutator] Initialize the config information to “safe” default values. An irrecoverable error will result if an uninitialized configuration is used to create a grammar.
Return value: Always succeeds. The return value is indeterminate.
[Accessor] Error codes are usually kept in the base grammar,
which leaves
marpa_g_new()
no place to put
its error code on failure.
Objects of
the Marpa_Config
class provide such a place.
p_error_string is reserved for use by
the internals.
Applications should set it to NULL
.
Return value: The error code in config.
Always succeeds, so that
marpa_c_error()
never requires an error code
for itself.
Next: Recognizer methods, Previous: Configuration methods, Up: Top [Contents][Index]
• Grammar overview: | ||
• Grammar constructor: | ||
• Grammar reference counting: | ||
• Symbol methods: | ||
• Rule methods: | ||
• Sequence methods: | ||
• Rank methods: | ||
• Grammar precomputation: |
Next: Grammar constructor, Previous: Grammar methods, Up: Grammar methods [Contents][Index]
An archetypal application has a grammar.
To create a grammar, use the
marpa_g_new()
method.
When a grammar is no longer in use, its memory can be freed
using the
marpa_g_unref()
method.
To be precomputed,
a grammar must have one or more symbols.
To create symbols, use the
marpa_g_symbol_new()
method.
To be precomputed,
a grammar must have one or more rules.
To create rules, use the
marpa_g_rule_new()
and
marpa_g_sequence_new()
methods.
For non-trivial parsing,
one or more of the symbols must be terminals.
To mark a symbol as a terminal,
use the
marpa_g_symbol_is_terminal_set()
method.
To be precomputed,
a grammar must have exactly one start symbol.
To mark a symbol as the start symbol,
use the
marpa_g_start_symbol_set()
method.
Before parsing with a grammar, it must be precomputed.
To precompute a grammar,
use the
marpa_g_precompute()
method.
Next: Grammar reference counting, Previous: Grammar overview, Up: Grammar methods [Contents][Index]
[Constructor] Creates a new grammar time object. The returned grammar object is not yet precomputed, and will have no symbols and rules. Its reference count will be 1.
Unless the application calls
marpa_c_error()
Libmarpa will not reference the location
pointed to by the configuration
argument after
marpa_g_new()
returns.
(See marpa_c_error().)
The configuration argument may be NULL
,
but if it is,
there will be no way to determine
the error code on failure.
Return value: On success, the grammar object.
On hard failure, NULL
.
Also on hard failure,
if the configuration argument is not NULL
,
the error code is set in configuration.
The error code may be accessed using
marpa_c_error()
.
[Mutator] It is recommended that this call be made immediately after the grammar constructor. It turns off a deprecated feature.
The
marpa_g_force_valued()
forces all the
symbols in a grammar to be “valued”.
The opposite of a valued symbol is one about whose value
you do not care.
This distinction has been made in the past in hope
of gaining efficiencies at evaluation time.
Current thinking is that the gains do not repay the extra
complexity.
Return value: On success, a non-negative integer, whose value is otherwise indeterminate. On failure, -2.
Next: Symbol methods, Previous: Grammar constructor, Up: Grammar methods [Contents][Index]
[Mutator] Increases the reference count of g by 1. Not needed by most applications.
Return value:
On success, g.
On hard failure, NULL
.
[Destructor] Decreases the reference count by 1, destroying g once the reference count reaches zero.
Next: Rule methods, Previous: Grammar reference counting, Up: Grammar methods [Contents][Index]
[Accessor] When successful, returns the ID of the start symbol.
Soft fails, if there is no start symbol.
The start symbol is set by the
marpa_g_start_symbol_set()
call.
Return value: On success, the ID of the start symbol, which is always a non-negative number. On soft failure, -1. On hard failure, -2.
[Mutator] When successful, sets the start symbol of grammar g to symbol sym_id. Soft fails if sym_id is well-formed (a non-negative integer), but a symbol with that ID does not exist.
Return value: On success, sym_id, which will always be a non-negative number. On soft failure, -1. On hard failure, -2.
[Accessor] Return value: On success, the numerically largest symbol ID of g. On hard failure, -2.
[Accessor] A symbol is accessible if it can be reached from the start symbol. Soft fails if sym_id is well-formed (a non-negative integer), but a symbol with that ID does not exist. A common hard failure is calling this method with a grammar that is not precomputed.
Return value: On success, 1 if symbol sym_id is accessible, 0 if not. On soft failure, -1. On hard failure, -2.
[Accessor] A symbol is nullable if it sometimes produces the empty string. A nulling symbol is always a nullable symbol, but not all nullable symbols are nulling symbols. Soft fails if sym_id is well-formed (a non-negative integer), but a symbol with that ID does not exist. A common hard failure is calling this method with a grammar that is not precomputed.
Return value: On success, 1 if symbol sym_id is nullable, 0 if not. On soft failure, -1. On hard failure, -2.
[Accessor] A symbol is nulling if it always produces the empty string. Soft fails if sym_id is well-formed (a non-negative integer), but a symbol with that ID does not exist. A common hard failure is calling this method with a grammar that is not precomputed.
Return value: On success, 1 if symbol sym_id is nulling, 0 if not. On soft failure, -1. On hard failure, -2.
[Accessor] A symbol is productive if it can produce a string of terminals. All nullable symbols are considered productive. Soft fails if sym_id is well-formed (a non-negative integer), but a symbol with that ID does not exist. A common hard failure is calling this method with a grammar that is not precomputed.
Return value: On success, 1 if symbol sym_id is productive, 0 if not. On soft failure, -1. On hard failure, -2.
[Accessor] On success, if sym_id is the start symbol, returns 1. On success, if sym_id is not the start symbol, returns 0. On success, if no start symbol has been set, returns 0. is the start symbol.
Soft fails if sym_id is well-formed (a non-negative integer), but a symbol with that ID does not exist.
Return value: On success, 1 or 0. On soft failure, -1. On hard failure, -2.
[Accessor] On succcess, returns the “terminal status” of a sym_id.
The terminal status is 1 if sym_id is a terminal,
0 otherwise.
To be used as an input symbol
in the
marpa_r_alternative()
method,
a symbol must be a terminal.
By default, a symbol is a terminal if and only if it
does not appear on the LHS of any rule.
The terminal status can be set explicitly with the
marpa_g_symbol_is_terminal_set()
method.
See marpa_g_symbol_is_terminal_set().
Soft fails if sym_id is well-formed (a non-negative integer), but a symbol with that ID does not exist.
Return value: On success, 1 or 0. On soft failure, -1. On hard failure, -2.
[Mutator] Sets the “terminal status” of a symbol.
This function flags symbol sym_id as a terminal if
value is 1,
or flags it as a non-terminal if value is 0.
To be used as an input symbol
in the
marpa_r_alternative()
method,
a symbol must be a terminal.
On success, this method returns value.
Once set to a value with this method,
the terminal status of a symbol is “locked” at that value.
A subsequent call to this method
that attempts
to change the terminal status
of sym_id to a value different from its current one
will hard fail
with error code MARPA_ERR_TERMINAL_IS_LOCKED
.
Other hard failures include
when value is not 0 or 1;
and
when the grammar g is precomputed.
By default, a symbol is a terminal if and only if it does not appear on the LHS of any rule. An attempt to flag a nulling symbol as a terminal will cause a failure, but this is not necessarily detected before precomputation.
Return value: On success, value, which will be 1 or 0. On soft failure, -1. On hard failure, -2.
[Mutator] When successful, creates a new symbol in grammar g.
Return value: On success, the ID of the new symbol; which will be a non-negative integer. On hard failure, -2.
Next: Sequence methods, Previous: Symbol methods, Up: Grammar methods [Contents][Index]
[Accessor] Return value: On success, the numerically largest rule ID of g. On hard failure, -2.
[Accessor] A rule is accessible if it can be reached from the start symbol. A rule is accessible if and only if its LHS symbol is accessible. The start rule is always an accessible rule.
Soft fails if rule_id is well-formed (a non-negative integer), but a rule with that ID does not exist. A common hard failure is calling this method with a grammar that is not precomputed.
Return value: On success 1 or 0: 1 if rule with ID rule_id is accessible, 0 if not. On soft failure, -1. On hard failure, -2.
[Accessor] A rule is nullable if it sometimes produces the empty string. A nulling rule is always a nullable rule, but not all nullable rules are nulling rules.
Soft fails if rule_id is well-formed (a non-negative integer), but a rule with that ID does not exist. A common hard failure is calling this method with a grammar that is not precomputed.
Return value: On success 1 or 0: 1 if the rule with ID rule_id is nullable, 0 if not. On soft failure, -1. On hard failure, -2.
[Accessor] A rule is nulling if it always produces the empty string.
Soft fails if rule_id is well-formed (a non-negative integer), but a rule with that ID does not exist. A common hard failure is calling this method with a grammar that is not precomputed.
Return value: On success 1 or 0: 1 if the rule with ID rule_id is nulling, 0 if not. On soft failure, -1. On hard failure, -2.
[Accessor] A rule is a loop rule if it non-trivially produces the string of length one which consists only of its LHS symbol. Such a derivation takes the parse back to where it started, hence the term “loop”. “Non-trivially” means the zero-step derivation does not count — the derivation must have at least one step.
The presence of a loop rule makes a grammar infinitely ambiguous, and applications will typically want to treat them as fatal errors. But nothing forces an application to do this, and Marpa will successfully parse and evaluate grammars with loop rules.
Soft fails if rule_id is well-formed (a non-negative integer), but a rule with that ID does not exist. A common hard failure is calling this method with a grammar that is not precomputed.
Return value: On success 1 or 0: 1 if the rule with ID rule_id is a loop rule, 0 if not. On soft failure, -1. On hard failure, -2.
[Accessor] A rule is productive if it can produce a string of terminals. A rule is productive if and only if all the symbols on its RHS are productive. The empty string counts as a string of terminals, so that a nullable rule is always a productive rule. For that same reason, an empty rule is considered productive.
Soft fails if rule_id is well-formed (a non-negative integer), but a rule with that ID does not exist. A common hard failure is calling this method with a grammar that is not precomputed.
Return value: On success 1 or 0: 1 if the rule with ID rule_id is productive, 0 if not. On soft failure, -1. On hard failure, -2.
[Accessor] The length of a rule is the number of symbols on its RHS.
Soft fails if rule_id is well-formed (a non-negative integer), but a rule with that ID does not exist.
Return value: On success, the length of the rule with ID rule_id. On soft failure, -1. On hard failure, -2.
[Accessor] Soft fails if rule_id is well-formed (a non-negative integer), but a rule with that ID does not exist.
Return value: On success, the ID of the LHS symbol of the rule with ID rule_id. On soft failure, -1. On hard failure, -2.
[Mutator] On success, creates a new external BNF rule in grammar g.
The ID of the new rule will be a non-negative integer,
which will be unique to that rule.
In addition to BNF rules, Marpa also allows sequence rules,
which are created by
the
marpa_g_sequence_new()
method.
See marpa_g_sequence_new().
Sequence rules and BNF rules are both rules: They share the same series of rule IDs, and are accessed and manipulated by the same methods, with the only differences being as noted in the descriptions of those methods.
The LHS symbol is lhs_id, and there are length symbols on the RHS. The RHS symbols are in an array pointed to by rhs_ids.
Possible hard failures, with their error codes, include:
MARPA_ERR_SEQUENCE_LHS_NOT_UNIQUE
: The LHS symbol is the same
as that of a sequence rule.
MARPA_ERR_DUPLICATE_RULE
: The new rule would duplicate another BNF
rule.
Another BNF rule is considered the duplicate of the new one,
if its LHS symbol is the same as symbol lhs_id,
if its length is the same as length,
and if its RHS symbols match one for one those
in the array of symbols rhs_ids.
Return value: On success, the ID of the new external rule. On hard failure, -2.
[Accessor] When successful, returns the ID of the symbol at index ix in the RHS of the rule with ID rule_id. The indexing of RHS symbols is zero-based.
Soft fails if rule_id is well-formed (a non-negative integer), but a rule with that ID does not exist.
A common hard failure is for ix not to be a valid index of the RHS. This happens if ix is less than zero, or or if ix is greater than or equal to the length of the rule.
Return value: On success, a symbol ID, which is always non-negative. On soft failure, -1. On hard failure, -2.
Next: Rank methods, Previous: Rule methods, Up: Grammar methods [Contents][Index]
[Accessor] When successful, returns
Does not distinguish sequence rules without proper
separation from non-sequence rules.
That is,
does not distinguish an unset proper separation flag
from a
proper separation flag which value is undefined
because rule_id is the ID of a BNF rule.
Applications which want to determine whether
or not a rule is a sequence rule
can use
marpa_g_sequence_min()
to do this.
See marpa_g_sequence_min().
Soft fails if rule_id is well-formed (a non-negative integer), but a rule with that ID does not exist.
Return value: On success, 1 or 0. On soft failure, -1. On hard failure, -2.
[Accessor] On success, returns the mininum length of a sequence rule. Soft fails if a rule with ID rule_id exists, but is not a sequence rule. This soft failure can used to test whether or not a rule is a sequence rule.
Hard fails irrecoverably if rule_id is not well-formed (a non-negative number). Also, hard fails irrecoverably if no rule with ID rule_id exists, even when rule_id is well formed. Note that, in its handling of the non-existence of a rule for its rule argument, this method differs from many of the other grammar methods. Grammar methods which take a rule ID argument more often treat the non-existence of rule for a well-formed rule ID as a soft, recoverable, failure.
Return value: On success, the minimum length of the sequence rule with ID rule_id, which is always non-negative. On soft failure, -1. On hard failure, -2.
[Mutator] When successful, adds a new sequence rule to grammar g, and return its ID. The ID of the sequence rule will be a non-negative integer, which is unique to that rule. All rules are numbered in the same series, so that a BNF rule will never have the same rule ID as a sequence rule, and vice versa.
Sequence rules are “sugar” — their presence in the Libmarpa interface does not extend its power. Every Libmarpa grammar which can be written using sequence rules can be rewritten as a grammar without sequence rules.
The LHS of the sequence is lhs_id, and the item to be repeated on the RHS of the sequence is rhs_id. The sequence must be repeated at least min times, where min is 0 or 1. If separator_id is non-negative, it is a separator symbol.
The LHS symbol cannot be the LHS of any other rule,
whether a BNF rule or a sequence rule.
On an attempt to create an sequence rule with a duplicate
LHS,
this method hard fails,
with an error code of
MARPA_ERR_SEQUENCE_LHS_NOT_UNIQUE
.
The sequence RHS, or item, is restricted to a single symbol, and that symbol cannot be nullable. If separator_id is a symbol, it also cannot be a nullable symbol. Nullables on the RHS of sequence rules are prohibited because it is not completely clear what an application intends when it asks for a sequence of items, some of which are nullable — the most natural interpretation of this usually results in a highly ambiguous grammar.
Libmarpa allows highly ambiquous grammars and a programmer who wants a grammar with sequences containing nullable items or separators can can write that grammar using BNF rules. The use of BNF rules make it clearer that ambiguity is what the programmer intended, and allows the programmer more flexibility.
If flags & MARPA_PROPER_SEPARATION
is non-zero,
separation is “proper”, that is,
a trailing separator is not allowed.
The term
proper
is based on the idea that
properly-speaking, separators should actually separate items.
Proper separation has no effect at the Libmarpa level —
it is tracked as a convenience for the higher-level interfaces to
Libmarpa,
which may want to offer the ability to
discard separators in the semantics.
(Some higher-level interfaces, in fact, may choose to discard separation by default.)
At the Libmarpa level, sequences always “keep
separators”.
Return value: On success, the ID of the newly added sequence rule, which is always non-negative. On hard failure, -2.
[Accessor] On success, returns the symbol ID of the separator of the sequence rule with ID rule_id. Soft fails if there is no separator. The causes of hard failure include rule_id not being well-formed; rule_id not being the ID of a rule which exists; and rule_id not being the ID a sequence rule.
Return value: On success, a symbol ID, which is always non-negative. On soft failure, -1. On hard failure, -2.
[Accessor] On success, returns a boolean whose value is 1 iff the symbol with ID sym_id is counted. A symbol is counted iff
Soft fails iff sym_id is well-formed (a non-negative integer), but a symbol with that ID does not exist.
Return value: On success, a boolean. On soft failure, -1. On hard failure, -2.
Next: Grammar precomputation, Previous: Sequence methods, Up: Grammar methods [Contents][Index]
[Accessor] When successful, returns the rank of the rule with ID rule_id. When a rule is created, its rank is initialized to the default rank of the grammar. The default rank of the grammar is 0.
Return value:
On success, returns a rule rank,
and sets the error code to
MARPA_ERR_NONE
.
The rule rank is an integer.
On hard failure, returns -2,
and sets the error code to an appropriate
value, which will never be
MARPA_ERR_NONE
.
Note that -2 is a valid rule rank,
so that when -2 is returned,
the error code is the only way to distinguish
success from failure.
The error code can be determined using
marpa_g_error()
.
See marpa_g_error().
[Mutator] When successful, sets the rank of the rule with ID rule_id to rank and returns rank.
Return value:
On success, returns rank,
which will be an integer,
and sets the error code to
MARPA_ERR_NONE
.
On hard failure, returns -2,
and sets the error code to an appropriate
value, which will never be
MARPA_ERR_NONE
.
Note that -2 is a valid rule rank,
so that when -2 is returned,
the error code is the only way to distinguish
success from failure.
The error code can be determined using
marpa_g_error()
.
See marpa_g_error().
[Accessor] On success, returns a boolean whose value is 1 iff “null ranks high” is set in the rule with ID rule_id. When a rule is created, it has “null ranks high” set.
For more on the
“null ranks high” setting, read the description of
marpa_g_rule_null_high_set()
.
See marpa_g_rule_null_high_set().
Soft fails iff rule_id is well-formed (a non-negative integer), but a rule with that ID does not exist.
Return value: On success, a boolean. On soft failure, -1. On hard failure, -2.
[Mutator] On success,
The “null ranks high” setting affects the ranking of rules with properly nullable symbols on their right hand side. If a rule has properly nullable symbols on its RHS, each instance in which it appears in a parse will have a pattern of nulled and non-nulled symbols. Such a pattern is called a “null variant”.
If the “null ranks high” is set, nulled symbols rank high. If the “null ranks high” is unset is the default), nulled symbols rank low. Ranking of a null variants is done from left-to-right.
Soft fails iff rule_id is well-formed (a non-negative integer), but a rule with that ID does not exist.
Hard fails if the grammar has been precomputed.
Return value: On success, a boolean. On soft failure, -1. On hard failure, -2.
Previous: Rank methods, Up: Grammar methods [Contents][Index]
[Accessor] On success, returns a boolean which is 1 iff
g has a cycle.
Cycles make a grammar infinitely ambiguous,
and are considered useless in current
practice.
Cycles make processing the grammar less
efficient, sometimes considerably so.
Applications will almost always want to treat cycles
as mistakes on the part of the writer of the grammar.
To determine which rules are in the cycle,
marpa_g_rule_is_loop()
can be used.
Return value: On success, a boolean. On hard failure, -2.
[Accessor] Return value: On success, a boolean which is 1 iff grammar g is precomputed. On hard failure, -2.
[Mutator] On success, and on fully recoverable hard failure, precomputes the grammar g. Precomputation involves running a series of grammar checks and “precomputing” some useful information which is kept internally to save repeated calculations. After precomputation, the grammar is “frozen” in many respects, and many grammar mutators which succeed before precomputation will cause hard failures after precomputation. Precomputation is necessary for a recognizer to be generated from a grammar.
When called, clears any events already in the event queue.
May return one or more events.
The types of event that this method may return
are
A MARPA_EVENT_LOOP_RULES
,
MARPA_EVENT_COUNTED_NULLABLE
,
MARPA_EVENT_NULLING_TERMINAL
.
All of these events occur only on failure.
Applications must be prepared for this method
to return additional events,
including events which occur on success.
Events may be queried using the
marpa_g_event()
method.
See marpa_g_event().
The fully recoverable hard failure is
MARPA_ERR_GRAMMAR_HAS_CYCLE
.
Recall that for fully recoverable hard failures
this method precomputes the grammar.
Most appplications, however, will want to treat
a grammar with cycles as if it were
a library-recoverable error.
A MARPA_ERR_GRAMMAR_HAS_CYCLE
error occurs
iff
a MARPA_EVENT_LOOP_RULES
event occurs.
For more details on cycles,
see marpa_g_has_cycle().
The error code MARPA_ERR_COUNTED_NULLABLE
is library-recoverable.
This failure occurs when a symbol on the RHS of a sequence rule is
nullable,
which Libmarpa does not allow in a grammar.
Error code MARPA_ERR_COUNTED_NULLABLE
occurs iff
one or more MARPA_EVENT_COUNTED_NULLABLE
events occur.
There is one MARPA_EVENT_COUNTED_NULLABLE
event for every symbol
which is a nullable on the right hand side of a sequence
rule.
An application may use these events to inform the user
of the problematic symbols,
and this detail may help the user fix the grammar.
The error code item MARPA_ERR_NULLING_TERMINAL
is library-recoverable.
This failure occurs when a nulling
symbol is also flagged as a terminal.
Since terminals cannot be of zero length, this is a logical
impossibility,
and Libmarpa does not allow nulling terminals in a grammar.
Error code item MARPA_ERR_NULLING_TERMINAL
occurs iff
one or more MARPA_EVENT_NULLING_TERMINAL
events occur.
There is one MARPA_EVENT_NULLING_TERMINAL
events
for every nulling terminal in the grammar.
An application may use these events to inform the user
of the problematic symbols,
and this detail may help the user fix the grammar.
Among the other error codes which may case this method to fail are the following:
MARPA_ERR_NO_RULES
: The grammar has no rules.
MARPA_ERR_NO_START_SYMBOL
: No start symbol was specified.
MARPA_ERR_INVALID_START_SYMBOL
: A start symbol ID was specified, but it
is not the ID of a valid symbol.
MARPA_ERR_START_NOT_LHS
: The start symbol is not on the LHS of any rule.
MARPA_ERR_UNPRODUCTIVE_START
: The start symbol is not productive.
More details of these can be found under the description of the appropriate code. See External error codes.
Return value: On success, a non-negative number,
whose value is otherwise indeterminate.
On hard failure, -2.
For the error code MARPA_ERR_GRAMMAR_HAS_CYCLE
,
the hard failure is fully recoverable.
For the error codes MARPA_ERR_COUNTED_NULLABLE
and MARPA_ERR_NULLING_TERMINAL
,
the hard failure is library-recoverable.
Next: Progress reports, Previous: Grammar methods, Up: Top [Contents][Index]
• Recognizer overview: | ||
• Creating a new recognizer: | ||
• Recognizer reference counting: | ||
• Recognizer life cycle mutators: | ||
• Location accessors: | ||
• Other parse status methods: |
Next: Creating a new recognizer, Previous: Recognizer methods, Up: Recognizer methods [Contents][Index]
An archetypal application uses a recognizer to read input.
To create a recognizer, use the
marpa_r_new()
method.
When a recognizer is no longer in use, its memory can be freed
using the
marpa_r_unref()
method.
To make a recognizer ready for input,
use the
marpa_r_start_input()
method.
The recognizer starts with its current earleme
at location 0.
To read a token at the current earleme,
use the
marpa_r_alternative()
call.
To complete the processing of the current earleme,
and move forward to a new one,
use the
marpa_r_earleme_complete()
call.
Next: Recognizer reference counting, Previous: Recognizer overview, Up: Recognizer methods [Contents][Index]
[Constructor] On success, creates a new recognizer and increments the reference count of g, the base grammar, by one. In the new recognizer,
Return value:
On success, the newly created recognizer, which is never NULL
.
If g is not precomputed, or on other hard failure, NULL
.
Next: Recognizer life cycle mutators, Previous: Creating a new recognizer, Up: Recognizer methods [Contents][Index]
[Mutator] Increases the reference count by 1. This method is not needed by most applications.
Return value:
On success, the recognizer object, r, which is never NULL
.
On hard failure, NULL
.
[Destructor] Decreases the reference count by 1, destroying r once the reference count reaches zero. When r is destroyed, the reference count of its base grammar is decreased by one. If this takes the reference count of the base grammar to zero, the base grammar is also destroyed.
Next: Location accessors, Previous: Recognizer reference counting, Up: Recognizer methods [Contents][Index]
[Mutator] When successful, does the following:
MARPA_EVENT_EXHAUSTED
event.
See Exhaustion.
MARPA_EVENT_SYMBOL_NULLED
,
MARPA_EVENT_SYMBOL_PREDICTED
, or MARPA_EVENT_SYMBOL_EXPECTED
events.
See Events.
Return value: On success, a non-negative value, whose value is otherwise indeterminate. On hard failure, -2.
The token_id argument must be the symbol ID of a terminal. The value argument is an integer that represents the “value” of the token, and which should not be zero. The length argument is the length of the token, which must be greater than zero.
On success, does the following, where current is the value of the current earleme before the call and furthest is the value of the furthest earleme before the call:
current+length
.
max(current+length,furthest)
.
After recoverable failure, the following are the case:
Libmarpa allows tokens to be ambiguous. Two tokens are ambiguous if they end at the same earleme location. If two tokens are ambiguous, Libmarpa will attempt to produce all the parses that include either of them.
Libmarpa allows tokens to overlap. Let the notation t@s-e indicate that token t starts at earleme s and ends at earleme e. Let t1@s1-e1 and t2@s2-e2 be two tokens such that s1<=s2. We say that t1 and t2 overlap iff e1>s2.
The value argument is not used inside Libmarpa — it is simply stored to be returned by the valuator as a convenience for the application. In applications where the token’s actual value is not an integer, it is expected that the application will use value as a “virtual” value, perhaps finding the actual value by using value to index an array. Some applications may prefer to track token values on their own, perhaps based on the earleme location and token_id, instead of using Libmarpa’s token values.
A value of 0 does not cause a failure, but it is reserved for unvalued symbols, a now-deprecated feature. See Valued and unvalued symbols.
Hard fails irrecoverably with MARPA_ERR_DUPLICATE_TOKEN
if the token added would be a duplicate.
Two tokens are duplicates iff all of the following are true:
marpa_r_alternative()
attempts
to read them while at the same current earleme.
If a token was not accepted
because of its token ID,
hard fails with the MARPA_ERR_UNEXPECTED_TOKEN_ID
.
This hard failure is fully recoverable
so that, for example,
the application may
retry this method with different token IDs
until it succeeds.
These retries are efficient,
and are quite useable as a parsing
technique —
so much so we have given the technique a name:
the Ruby Slippers.
The Ruby Slippers are used in several
applications.
Return value: On success, MARPA_ERR_NONE
.
On failure, an error code other than MARPA_ERR_NONE
.
The hard failure for MARPA_ERR_UNEXPECTED_TOKEN_ID
is fully recoverable.
For the purposes of this method description, we define the following:
marpa_r_earleme_complete
.
marpa_r_earleme_complete
.
marpa_r_terminal_is_expected()
determines if a terminal is “expected”
at the current earleme.
See marpa_r_terminals_expected().
marpa_r_alternative()
to end at an earleme after the current
earleme.
An anticipated terminal will have length greater than one.
“Anticipated” terminals only occur if the application is using
an advanced model of input.
See Advanced input models.
On success, does the final processing for the current earleme, including the following:
current+1
.
current+1
.
marpa_r_earleme_complete()
.
MARPA_EVENT_SYMBOL_COMPLETED
,
MARPA_EVENT_SYMBOL_NULLED
, MARPA_EVENT_SYMBOL_PREDICTED
,
or MARPA_EVENT_SYMBOL_EXPECTED
events.
See Events.
MARPA_EVENT_EARLEY_ITEM_THRESHOLD
event.
Often, the application will want to treat this event
as if it were a library-recoverable
failure.
See marpa_r_earley_item_warning_threshold_set().
MARPA_EVENT_EXHAUSTED
event.
Exhaustion on success only occurs if no terminals
are expected at the current earleme after
the call to this method
(that is, at current+1
)
and no terminals are anticipated
after current+1
.
On hard failure
with the code MARPA_ERR_PARSE_EXHAUSTED
, does the following:
marpa_r_earleme_complete()
.
MARPA_EVENT_EXHAUSTED
event and no others.
MARPA_ERR_PARSE_EXHAUSTED
to be fully recoverable.
We note that exhaustion can occur when this method fails and when it succeeds. The distinction is that, on success, the call creates a new Earley set before becoming exhausted while, on failure, it becomes exhausted without creating a new Earley set.
Return value: On success, the number of events generated.
On hard failure, -2.
Hard failure with the code
MARPA_ERR_PARSE_EXHAUSTED
is fully recoverable.
Next: Other parse status methods, Previous: Recognizer life cycle mutators, Up: Recognizer methods [Contents][Index]
Marpa_Earleme
marpa_r_current_earleme (Marpa_Recognizer r)Return value: If input has started, the current earleme. If input has not started, -1. Always succeeds.
In the default, token-stream model, Earley set ID and earleme
are always equal, but this is not the case in other input
models.
(The ID of an Earley set ID is also called its ordinal.)
If there is no Earley set whose ID is
set_id,
marpa_r_earleme()
fails.
If set_id was negative,
the error code is set to
MARPA_ERR_INVALID_LOCATION
.
If set_id is greater than the ordinal
of the latest Earley set,
the error code is set to
MARPA_ERR_NO_EARLEY_SET_AT_LOCATION
.
At this writing, there is no method for
the inverse operation (conversion of an earleme to an Earley set
ID).
One consideration in writing
such a method is that not all earlemes correspond to Earley sets.
Applications that want to map earlemes
to Earley sets will have no trouble if they
are using the standard input model —
the Earley set
ID is always exactly equal to the earleme in that model.
For other applications
that want an earleme-to-ID mapping,
the most general method is create an ID-to-earleme
array using the
marpa_r_earleme()
method
and invert it.
Return value: On success, the earleme corresponding to Earley set set_id. On failure, -2.
Returns the integer value of earley_set.
For more details, see
the description of
marpa_r_earley_set_values()
.
Return value: On success, the value of earley_set. On failure, -2.
If p_value is non-zero, sets the location pointed to by p_value to the integer value of the Earley set. Similarly, if p_pvalue is non-zero, sets the location pointed to by p_pvalue to the pointer value of the Earley set.
The “value” and “pointer” of an Earley set are an arbitrary integer and an arbitrary pointer that the application can use for its own purposes. In character-per-earleme input models, for example, the integer can be the codepoint of the current character. In a traditional token-per-earleme input model, they could be used to indicate the string value of the token – the pointer could point to the start of the string, and the integer could indicate its length.
The Earley set value and pointer can be set using
the
marpa_r_latest_earley_set_values_set()
method.
The Earley set integer value defaults to -1,
and the pointer value defaults to NULL
.
Return value: On success, returns a non-negative integer. On failure, returns -2.
unsigned int
marpa_r_furthest_earleme (Marpa_Recognizer r)Always returns the furthest earleme.
Return value: On success, the furthest earleme. Always succeeds.
This method returns the Earley set ID (ordinal) of the latest Earley set.
Applications that want the
value of the latest earleme can convert
this value using
the
marpa_r_earleme()
method.
Return value: On success, the ID of the latest Earley set. Always succeeds.
Sets the integer value of the latest Earley set.
For more details, see
the description of
marpa_r_latest_earley_set_values_set()
.
Return value: On success, the new value of earley_set. On failure, -2.
Sets the integer and pointer value of the latest Earley set.
For more about the “integer value” and “pointer value”
of an Earley set,
see the description of the
marpa_r_earley_set_values()
method.
Return value: On success, returns a non-negative integer. On failure, returns -2.
Previous: Location accessors, Up: Recognizer methods [Contents][Index]
Returns the Earley item warning threshold. See marpa_r_earley_item_warning_threshold_set().
Return value: The Earley item warning threshold. Always succeeds.
[Mutator] On success, sets
the Earley item warning threshold.
The
Earley item warning threshold
is a number that is compared with
the count of Earley items in each Earley set.
When it is matched or exceeded,
a MARPA_EVENT_EARLEY_ITEM_THRESHOLD
event is created.
See MARPA_EVENT_EARLEY_ITEM_THRESHOLD.
If threshold is zero or less, an unlimited number of Earley items will be allowed without warning. This will rarely be what the user wants.
By default, Libmarpa calculates a value based on the grammar. The formula Libmarpa uses is the result of some experience, and most applications will be happy with it.
What should be done when the threshold is exceeded, depends on the application,
but exceeding the threshold means that it is very likely
that the time and space resources consumed by
the parse will prove excessive.
This is often a sign of a bug in the grammar.
Applications often will want to smoothly shut down
the parse,
in effect treating
the MARPA_EVENT_EARLEY_ITEM_THRESHOLD
event
as equivalent to library-recoverable hard failure.
Return value: The value that the Earley item warning threshold has after the method call is finished. Always succeeds.
A parser is “exhausted” if it cannot accept any more input. Both successful and failed parses can be exhausted. In many grammars, the parse is always exhausted as soon as it succeeds. Good parses may also exist at earlemes prior to the current one.
Return value: 1 if the parser is exhausted, 0 otherwise. Always succeeds.
Returns a list of the ID’s of the symbols
that are acceptable as tokens
at the current earleme.
buffer is expected to be large enough to hold
the result.
This is guaranteed to be the case if the buffer
is large enough to hold a number of
Marpa_Symbol_ID
’s that
is greater than or equal to the number of symbols
in the grammar.
Return value: On success, the number of Marpa_Symbol_ID
’s
in buffer.
On failure, -2.
Return values on success: If symbol_id is the ID of a valid terminal symbol that is expected at the current earleme, a number greater than zero. If symbol_id is the ID of a valid terminal symbol that is not expected at the current earleme, or if symbol_id is the ID of a valid symbol that is not a terminal, zero.
Failure cases: Returns -2 on failure. It is a failure if symbol_id is not the ID of a valid symbol.
Next: Bocage methods, Previous: Recognizer methods, Up: Top [Contents][Index]
An important advantage of the Marpa algorithm is the ability to easily get full information about the state of the parse.
To start a progress report,
use the
marpa_r_progress_report_start()
command.
Only one progress report can be in use at any one time.
To get the information in a progress report,
it is necessary to step through the progress report
items.
To get the data for the current progress report item,
and advance to the next one,
use the
marpa_r_progress_item()
method.
To destroy a progress report,
freeing the memory it uses,
call the
marpa_r_progress_report_finish()
method.
Resets the progress report.
Assumes a report of the progress has already been initialized
at some Earley set
for recognizer r,
with
marpa_r_progress_report_start()
.
The reset progress report will
be positioned before its first item.
Return value: On success, a non-negative value. On failure, -2.
Initializes a report of the progress at Earley set set_id for recognizer r. If a progress report already exists, it is destroyed and its memory is freed. Initially, the progress report is positioned before its first item.
If no Earley set with ID
set_id exists,
marpa_r_progress_report_start()
fails.
The error code is MARPA_ERR_INVALID_LOCATION
if set_id
is negative.
The error code is MARPA_ERR_NO_EARLEY_SET_AT_LOCATION
if set_id is greater than the ID of
the latest Earley set.
Return value: On success, the number of report items available. If the recognizer has not been started; if set_id does not exist; or on other failure, -2.
Destroys the report of the progress at Earley set set_id for recognizer r, freeing the memory and other resources. It is often not necessary to call this method. Any previously existing progress report is destroyed automatically whenever a new progress report is started, and when the recognizer is destroyed.
Return value: -2 if no progress report has been started, or on other failure. On success, a non-negative value.
This method allows access to the data
for the next item of a
progress report.
If there are no more progress report items,
it returns -1 as a termination indicator
and sets the error code to MARPA_ERR_PROGRESS_REPORT_EXHAUSTED
.
Either the termination indicator,
or the item count returned by
marpa_r_progress_report_start()
,
can be used to determine when the last
item has been seen.
On success, the dot position is returned in the location pointed to by the position argument, and the origin is returned in the location pointed to by the origin argument. On failure, the locations pointed to by the position and origin arguments are unchanged.
Return value: On success, the rule ID of
the next progress report item.
If there are no more progress report items, -1.
If either the position or the origin
argument is NULL
,
or on other failure, -2.
Next: Ordering methods, Previous: Progress reports, Up: Top [Contents][Index]
• Bocage overview: | ||
• Bocage constructor: | ||
• Bocage reference counting: | ||
• Bocage accessor: |
Next: Bocage constructor, Previous: Bocage methods, Up: Bocage methods [Contents][Index]
A bocage is structure containing the full set of parses found by processing the input according to the grammar. The bocage structure is new with Libmarpa, but is very similar in purpose to the more familar parse forests.
To create a bocage, use the
marpa_b_new()
method.
When a bocage is no longer in use, its memory can be freed
using the
marpa_b_unref()
method.
Next: Bocage reference counting, Previous: Bocage overview, Up: Bocage methods [Contents][Index]
Creates a new bocage object, with a reference count of 1. The reference count of its parent recognizer object, r, is increased by 1. If earley_set_ID is -1, the Earley set at the current earleme is used, if there is one.
If earley_set_ID is -1
and there is no Earley set at the current earleme;
or if earley_set_ID is -1
and there is no parse ending at Earley set earley_set_ID,
marpa_b_new()
fails
and the error code is set to
MARPA_ERR_NO_PARSE
.
Success return value: On success, the new bocage object.
On failure, NULL
.
Next: Bocage accessor, Previous: Bocage constructor, Up: Bocage methods [Contents][Index]
Increases the reference count by 1. Not needed by most applications.
Return value:
On success, b.
On failure, NULL
.
Decreases the reference count by 1, destroying b once the reference count reaches zero. When b is destroyed, the reference count of its parent recognizer is decreased by 1. If this takes the reference count of the parent recognizer to zero, it too is destroyed. If the parent recognizer is destroyed, the reference count of its base grammar is decreased by 1. If this takes the reference count of the base grammar to zero, it too is destroyed.
Previous: Bocage reference counting, Up: Bocage methods [Contents][Index]
Returns an ambiguity metric. The metric is 1 is the parse is unambiguous. If the metric is 2 or greater, the parse is ambiguous. It was originally intended to have values greater than 2 be an cheaply computed estimate of the degree of ambiguity, but a satisfactory scheme for this has yet to be implemented.
Return value on success: 1 if the bocage is not for an ambiguous parse; 2 or greater if the bocage is for an ambiguous parse.
Failures: On failure, -2.
Return value on success: A number greater than or equal to 1 if the bocage is for a null parse; otherwise, 0.
Failures: On failure, -2.
Next: Tree methods, Previous: Bocage methods, Up: Top [Contents][Index]
• Ordering overview: | ||
• Ordering constructor: | ||
• Ordering reference counting: | ||
• Order accessor: | ||
• Non-default ordering: |
Next: Ordering constructor, Previous: Ordering methods, Up: Ordering methods [Contents][Index]
Before iterating the parses in the bocage,
they must be ordered.
To create an ordering, use the
marpa_o_new()
method.
When an ordering is no longer in use, its memory can be freed
using the
marpa_o_unref()
method.
An ordering is frozen once the first tree iterator is created using it. A frozen ordering cannot be changed.
As of this writing, the only methods to order parses are internal and undocumented. This is expected to change.
Next: Ordering reference counting, Previous: Ordering overview, Up: Ordering methods [Contents][Index]
Creates a new ordering object, with a reference count of 1. The reference count of its parent bocage object, b, is increased by 1.
Return value: On success, the new ordering object.
On failure, NULL
.
Next: Order accessor, Previous: Ordering constructor, Up: Ordering methods [Contents][Index]
Increases the reference count by 1. Not needed by most applications.
Return value:
On success, o.
On failure, NULL
.
Decreases the reference count by 1, destroying o once the reference count reaches zero. Beginning with o’s parent bocage, Libmarpa then proceeds up the chain of parent objects. Every time a child is destroyed, the reference count of its parent is decreased by 1. Every time the reference count of an object is decreased by 1, if that reference count is now zero, that object is destroyed. Libmarpa follows this chain of decrements and destructions as required, all the way back to the base grammar, if necessary.
Next: Non-default ordering, Previous: Ordering reference counting, Up: Ordering methods [Contents][Index]
Returns an ambiguity metric. The metric is 1 is the parse is unambiguous. If the metric is 2 or greater, the parse is ambiguous. It was originally intended to have values greater than 2 be an cheaply computed estimate of the degree of ambiguity, but a satisfactory scheme for this has yet to be implemented.
If the ordering is not already frozen,
it will be frozen on return from
marpa_o_ambiguity_metric()
.
marpa_o_ambiguity_metric()
is considered an “accessor”,
because it is assumed that the ordering is frozen
when
marpa_o_ambiguity_metric()
is called.
Return value on success: 1 if the ordering is not for an ambiguous parse; 2 or greater if the ordering is for an ambiguous parse.
Failures: On failure, -2.
Return value on success: A number greater than or equal to 1 if the ordering is for a null parse; otherwise, 0.
Failures: On failure, -2.
Previous: Order accessor, Up: Ordering methods [Contents][Index]
These methods, respectively, set and query the “high rank only” flag of ordering o. A flag of 1 indicates that, when ranking, all choices should be discarded except those of the highest rank. A flag of 0 indicates that no choices should be discarded on the basis of their rank.
A value of 1 is the default.
The value of the “high rank only” flag has no effect
unless ranking has been turned on using the
marpa_o_rank()
method.
Return value: On success, the value of the “high rank only” flag after the call. On failure, -2.
By default, the ordering of parse trees is arbitrary. This method causes the ordering to be ranked according to the ranks of symbols and rules, the “null ranks high” flags of the rules, and the “high rank only” flag of the ordering. Once this method returns, the ordering is frozen.
Return value: On success, a non-negative value. On failure, -2.
Next: Value methods, Previous: Ordering methods, Up: Top [Contents][Index]
• Tree overview: | ||
• Tree constructor: | ||
• Tree reference counting: | ||
• Tree iteration: |
Next: Tree constructor, Previous: Tree methods, Up: Tree methods [Contents][Index]
Once the bocage has an ordering, the parses trees can be iterated. Marpa’s parse tree iterators iterate the parse trees contained in a bocage object. In Libmarpa, “parse tree iterators” are usually just called trees.
To create a tree, use the
marpa_t_new()
method.
A newly created tree iterator is positioned before the first parse tree.
When a tree iterator is no longer in use, its memory can be freed
using the
marpa_t_unref()
method.
To position a newly created tree iterator at the first parse tree,
use the
marpa_t_next()
method.
Once the tree iterator is positioned at a parse tree,
the same
marpa_t_next()
method is used
to position it to the next parse tree.
Next: Tree reference counting, Previous: Tree overview, Up: Tree methods [Contents][Index]
Creates a new tree iterator, with a reference count of 1. The reference count of its parent ordering object, o, is increased by 1.
When initialized, a tree iterator is positioned
before the first parse tree.
To position the tree iterator to the first parse,
the application must call
marpa_t_next()
.
Return value: On success, a newly created tree.
On failure, NULL
.
Next: Tree iteration, Previous: Tree constructor, Up: Tree methods [Contents][Index]
Increases the reference count by 1. Not needed by most applications.
Return value:
On success, t.
On failure, NULL
.
Decreases the reference count by 1, destroying t once the reference count reaches zero. Beginning with t’s parent ordering, Libmarpa then proceeds up the chain of parent objects. Every time a child is destroyed, the reference count of its parent is decreased by 1. Every time the reference count of an object is decreased by 1, if that reference count is now zero, that object is destroyed. Libmarpa follows this chain of decrements and destructions as required, all the way back to the base grammar, if necessary.
Previous: Tree reference counting, Up: Tree methods [Contents][Index]
Positions t at the next parse tree in the iteration. Tree iterators are initialized to the position before the first parse tree, so this method must be called before creating a valuator from a tree.
If a tree iterator is positioned after the last parse,
the tree is said to be “exhausted”.
A tree iterator for a bocage with no parse trees
is considered to be “exhausted” when initialized.
If the tree iterator is exhausted,
marpa_t_next()
returns -1 as a termination indicator,
and sets the error code to
MARPA_ERR_TREE_EXHAUSTED
.
Return value: On success, a non-negative value. If the tree iterator is exhausted, -1. On failure, -2.
The parse counter counts the number of parse trees traversed so far. The count includes the current iteration of the tree, so that a value of 0 indicates that the tree iterator is at its initialized position, before the first parse tree.
Return value: The number of parses traversed so far. Always succeeds.
Next: Events, Previous: Tree methods, Up: Top [Contents][Index]
Next: How to use the valuator, Previous: Value methods, Up: Value methods [Contents][Index]
The archetypal application needs
a value object (or
valuator) to produce
the value of the parse.
To create a valuator, use the
marpa_v_new()
method.
When a valuator is no longer in use, its memory can be freed
using the
marpa_v_unref()
method.
The application is required to maintain the stack,
and the application is also required to implement
most of the semantics, including the evaluation
of rules.
Libmarpa’s valuator provides instructions to
the application on how to manipulate the stack.
To iterate through this series of instructions,
use the
marpa_v_step()
method.
When successful, marpa_v_step()
returns the type
of step.
Most step types have values associated with them.
To access these values use the methods
described in the section Basic step accessors.
How to perform the steps is described in
the sections
How to use the valuator
and Stepping through the valuator.
Next: Advantages of step-driven valuation, Previous: Value overview, Up: Value methods [Contents][Index]
Libmarpa’s valuator provides the application with “steps”, which are instructions for stack manipulation. Libmarpa itself does not maintain a stack. This leaves the upper layer in total control of the stack and the values which are placed on it.
As example may make this clearer. Suppose the evalution is at a place in the parse tree where an addition is being performed. Libmarpa does not know that the operation is an addition. It will tell the application that rule number R is to be applied to the arguments at stack locations N and N+1, and that the result is to placed in stack location N.
In this system the application keeps track of the semantics for all rules, so it looks up rule R and determines that it is an addition. The application can do this by using R as an index into an array of callbacks, or by any other method it chooses. Let’s assume a callback implements the semantics for rule R. Libmarpa has told the application that two arguments are available for this operation, and that they are at locations N and N+1 in the stack. They might be the numbers 42 and 711. So the callback is called with its two arguments, and produces a return value, let’s say, 753. Libmarpa has told the application that the result belongs at location N in the stack, so the application writes 753 to location N.
Since Libmarpa knows nothing about the semantics, the operation for rule R could be string concatenation instead of addition. Or, if it is addition, it could allow for its arguments to be floating point or complex numbers. Since the application maintains the stack, it is up to the application whether the stack contains integers, strings, complex numbers, or polymorphic objects which are capable of being any of these things and more.
Next: Maintaining the stack, Previous: How to use the valuator, Up: Value methods [Contents][Index]
Step-driven valuation hides Libmarpa’s grammar rewrites from the application, and is quite efficient. Libmarpa knows which rules are sequences. Libmarpa optimizes stack manipulations based on this knowledge. Long sequences are very common in practical grammars. For these, the stack manipulations suggested by Libmarpa’s step-driven valuator will be significantly faster than the traditional stack evaluation algorithm.
Step-driven evalution has another advantage. To illustrate this, consider what is a very common case: The semantics are implemented in a higher-level language, using callbacks. If Libmarpa did not use step-driven valuation, it would need to provide for this case. But for generality, Libmarpa would have to deal in C callbacks. Therefore, a middle layer would have to create C language wrappers for the callbacks in the higher level language.
The implementation that results is this: The higher level language would need to wrap each callback in C. When calling Libmarpa, it would pass the wrappered callback. Libmarpa would then need to call the C language “wrappered” callback. Next, the wrapper would call the higher-level language callback. The return value, which would be data native to the higher-level language, would need to be passed to the C language wrapper, which will need to make arrangements for it to be based back to the higher-level language when appropriate.
A setup like this is not terribly efficient. And exception handling across language boundaries would be very tricky. But neither of these is the worst problem.
Callbacks are hard to debug. Wrappered callbacks are even worse. Calls made across language boundaries are harder yet to debug. In the system described above, by the time a return value is finally consumed, a language boundary will have been crossed four times.
How do Libmarpa users deal with difficulties like this? Usually, by doing the absolute minimum possible in the callbacks. A horrific debugging enviroment can become a manageable one if there is next to no code to be debugged. And this can be accomplished by doing as much as possible in pre- and post-processing.
In essence, callbacks force applications to do most of the programming via side effects. One need not be a functional programming purist to find this a very undesirable style of design to force on an application. But the ability to debug can make the difference between code that does work and code that does not. Unfairly or not, code is rarely considered well-designed when it does not work.
So, while step-driven valuation seems a roundabout approach, it is simpler and more direct than the likely alternatives. And there is something to be said for pushing semantics up to the higher levels — they can be expected to know more about it.
These advantages of step-driven valuation are strictly in the context of a low-level interface. The author is under no illusion that direct use of Libmarpa’s valuator will be found satisfactory by most Libmarpa users, even those using the C language. The author certainly avoids using step-driven valuation directly. Libmarpa’s valuator is intended to be used via an upper layer, one which does know about semantics.
Next: Valuator constructor, Previous: Advantages of step-driven valuation, Up: Value methods [Contents][Index]
This section discusses in detail the requirements for maintaining the stack. In some cases, such as implementation using a Perl array, fulfilling these requirements is trivial. Perl auto-extends its arrays, and initializes the element values, on every read or write. For the C programmer, things are not quite so easy.
In this section, we will assume a C90 or C99 standard-conformant C application. This assumption is convenient on two grounds. First, this will be the intended use for many readers. Second, standard-conformant C is a “worst case”. Any issue faced by a programmer of another environment is likely to also be one that must be solved by the C programmer.
Libmarpa often optimizes away unnecessary stack writes to stack locations. When it does so, it will not necessarily optimize away all reads to that stack location. This means that a location’s first access, as suggested by the Libmarpa step instructions, may be a read. This possibility requires a special awareness from the C programmer, as discussed in the sections Sizing the stack and Initializing locations in the stack.
In the discussions in this document, stack locations are non-negative integers. The bottom of the stack is location 0. In moving from the bottom of the stack to the top, the numbers increase. Stack location Y is said to be “greater” than stack location X if stack location Y is closer to the top of stack than location X, and therefore stack locations are considered greater or lesser if the integers that represent them are greater or lesser. Another way to state that a stack location Y is greater (lesser) than stack location X is to say that a stack location Y is later (earlier) than stack location X.
• Sizing the stack: | ||
• Initializing locations in the stack: |
Next: Initializing locations in the stack, Previous: Maintaining the stack, Up: Maintaining the stack [Contents][Index]
If an implementation applies Libmarpa’s step instructions literally, using a physical stack, it must make sure the stack is large enough. Specifically, the application must do the following
MARPA_STEP_TOKEN
steps,
ensure that location
marpa_v_result(v)
exists.
MARPA_STEP_NULLING_SYMBOL
steps,
ensure that location
marpa_v_result(v)
exists.
MARPA_STEP_RULE
steps,
ensure that stack locations from
marpa_v_arg_0(v)
to
marpa_v_arg_n(v)
exist.
Three aspects of these requirements deserve special mention.
First,
note that the requirement for a
MARPA_STEP_RULE
is that the application
size the stack to include the arguments to be
read.
Because stack writes may be optimized away,
an application,
when reading,
cannot assume
that the stack was
sized appropriately by a prior write.
The first access to a new stack location may be
a read.
Second,
note that there is no explicit requirement that
the application size the stack to include the
location for the result of the
MARPA_STEP_RULE
step.
An application is allowed to assume that
result will go into one of the locations
that were read.
Third, special note should be made of the requirement that location 0 exist. By convention, the parse result resides in location 0 of the stack. Because of potential optimizations, an application cannot assume that it will receive a Libmarpa step instruction that either reads from or writes to location 0.
Previous: Sizing the stack, Up: Maintaining the stack [Contents][Index]
Write optimizations also creates issues for implementations which require data to be initialized before reading. Every fully standard-conforming C application is such an implementation. Both C90 and C99 allow “trap values”, and therefore conforming applications must be prepared for an uninitialized location to contain one of those. Reading a trap value may cause an abend. (It is safe, in standard-conforming C, to write to a location containing a trap value.)
The requirement that locations be initialized before reading occurs in other implementations. Any implementation that has a “universe” of “safe” values, may require special precautions. The required precautions may amount to a need to initialize “uninitialized” values. A practical example might be an implementation that expects all locations to contain a pointer which it can safely indirect from. In such implementations, just as in standard-conformant C, every stack location needs to be initialized before being read.
Due to write optimizations, an application cannot rely on Libmarpa’s step instructions to initialize every stack location before its first read. One way to safely deal with the initialization of stack locations, is to do all of the following:
Applications which try to optimize out some of these initializations need to be aware that an application can never assume that activity in the stack is safely “beyond” an uninitialized location. Libmarpa steps often revisit earlier sections of the stack, and these revisits may include reads of previously unvisited stack locations.
Next: Valuator reference counting, Previous: Maintaining the stack, Up: Value methods [Contents][Index]
Creates a new valuator. The parent object of the new valuator will be the tree iterator t, and the reference count of the new valuator will be 1. The reference count of t is increased by 1.
The parent tree iterator is “paused”, so that the tree iterator cannot move on to a new parse tree until the valuator is destroyed. Many valuators of the same parse tree can exist at once. A tree iterator is “unpaused” when all of the valuators of that tree iterator are destroyed.
Return value: On success, the newly created valuator.
On failure, NULL
.
Next: Stepping through the valuator, Previous: Valuator constructor, Up: Value methods [Contents][Index]
Increases the reference count by 1. Not needed by most applications.
Return value:
On success, v.
On failure, NULL
.
Decreases the reference count by 1, destroying v once the reference count reaches zero. Beginning with v’s parent tree, Libmarpa then proceeds up the chain of parent objects. Every time a child is destroyed, the reference count of its parent is decreased by 1. Every time the reference count of an object is decreased by 1, if that reference count is now zero, that object is destroyed. Libmarpa follows this chain of decrements and destructions as required, all the way back to the base grammar, if necessary.
Next: Valuator steps by type, Previous: Valuator reference counting, Up: Value methods [Contents][Index]
This method “steps through” the valuator.
The return value is a Marpa_Step_Type
,
an integer which indicates the type of step.
How the application is expected to act on
each step is described below
(Valuator steps by type).
When the iteration through the steps is finished,
marpa_v_step()
returns MARPA_STEP_INACTIVE
.
Return value: On success, a Marpa_Step_Type
,
which always be a non-negative integer.
On failure, -2.
Next: Basic step accessors, Previous: Stepping through the valuator, Up: Value methods [Contents][Index]
The semantics of a rule should be performed.
The application can find the value of the rule’s
children in the stack locations from
marpa_v_arg_0(v)
to
marpa_v_arg_n(v)
.
The semantics for the rule whose ID is
marpa_v_rule(v)
should be executed
on these child values,
and the result placed in
marpa_v_result(v)
.
In the case of a MARPA_STEP_RULE
step,
the stack location of
marpa_v_result(v)
is guaranteed to
be equal to
marpa_v_arg_0(v)
.
The semantics of a non-null token should be performed.
The application’s value for the token whose ID is
marpa_v_token(v)
should be
placed in
stack location
marpa_v_result(v)
.
Its value according to Libmarpa will be in
marpa_v_token_value(v)
.
The semantics for a nulling symbol should be performed.
The ID of the symbol is
marpa_v_symbol(v)
and its value should
be placed in
stack location
marpa_v_result(v)
.
The valuator has gone through all of its steps
and is now inactive.
The value of the parse will be in stack location 0.
Because of optimizations,
it is possible for valuator to immediately
became inactive — MARPA_STEP_INACTIVE
could
be both the first and last step.
The valuator is new and has yet to go through any steps.
These step types are reserved for internal purposes.
Next: Other step accessors, Previous: Valuator steps by type, Up: Value methods [Contents][Index]
The basic step accessors are so called because their information is basic to the stack manipulation. The basic step accessors are implemented as macros. They always succeed.
For a MARPA_STEP_RULE
step,
returns the stack location where the value of first child
can be found.
For a MARPA_STEP_RULE
step,
returns the stack location where the value of the last child
can be found.
For MARPA_STEP_RULE
,
MARPA_STEP_TOKEN
,
and MARPA_STEP_NULLING_SYMBOL
steps,
returns the stack location where the result of the semantics
should be placed.
For the
MARPA_STEP_RULE
step,
returns the ID of the rule.
Returns the current step type:
MARPA_STEP_TOKEN
,
MARPA_STEP_RULE
,
etc.
Usually not needed since this is also the return value of
marpa_v_step()
.
For the MARPA_STEP_NULLING_SYMBOL
step,
returns the ID of the symbol.
The value returned is the same as that
returned by the
marpa_v_token()
macro.
For the MARPA_STEP_TOKEN
step,
returns the ID of the token.
The value returned is the same as that
returned by the
marpa_v_symbol()
macro.
For the MARPA_STEP_TOKEN
step,
returns the integer which is
(or which represents)
the value of the token.
Previous: Basic step accessors, Up: Value methods [Contents][Index]
This section contains the step accessors that are not basic to stack manipulation, but which provide other useful information about the parse. These step accessors are implemented as macros.
All of these accessors always succeed, but if called when they are irrelevant they return an unspecified value. In this context, an “unspecified value” is a value that is either -1 or the ID of a valid Earley set, but which is otherwise unpredictable.
Return value:
If the current step type is MARPA_STEP_RULE
,
the Earley Set ordinal where the rule ends.
If the current step type is
MARPA_STEP_TOKEN
or MARPA_STEP_NULLING_SYMBOL
,
the Earley Set ordinal where the symbol ends.
If the current step type is anything else, an unspecified value.
Return value:
If the current step type is MARPA_STEP_RULE
,
the Earley Set ordinal where the rule begins.
If the current step type is anything else, an unspecified value.
Return value:
If the current step type is MARPA_STEP_TOKEN
or MARPA_STEP_NULLING_SYMBOL
,
the Earley Set ordinal where the token begins.
If the current step type is anything else, an unspecified value.
Next: Error methods macros and codes, Previous: Value methods, Up: Top [Contents][Index]
• Events overview: | ||
• Basic event accessors: | ||
• Completion events: | ||
• Symbol nulled events: | ||
• Prediction events: | ||
• Symbol expected events: | ||
• Event codes: |
Next: Basic event accessors, Previous: Events, Up: Events [Contents][Index]
Events are generated by the
marpa_g_precompute()
,
marpa_r_earleme_complete()
,
and
marpa_r_start_input()
methods.
The methods are called event-active.
Event-active methods always clear all previous events,
so that after an event-active method the only events
available
will be those generated by that method.
Some Libmarpa methods clear the event queue. The user is expected to query events immediately after the method that generated them. We note especially that events are kept in the base grammar, so that multiple recognizers using the same base grammar overwrite each other’s events.
To find out how many events were generated by the last
event-active method,
use the
marpa_g_event_count()
method.
To query a specific event,
use the
marpa_g_event()
and
marpa_g_event_value()
methods.
In reading this chapter, we will need to be aware that it contains a mixture of grammar and recognizer methods.
Next: Completion events, Previous: Events overview, Up: Events [Contents][Index]
On success, the type of the ix’th event is returned and the data for the ix’th event is placed in the location pointed to by event.
Event indexes are in sequence.
Valid events will be in the range from 0 to n,
where n is one less than the event count.
The event count
can be queried using the
marpa_g_event_count()
method.
Return value: On success, the type of event ix. If there is no ix’th event, if ix is negative, or on other failure, -2. On failure, the locations pointed to by event are not changed.
Return value: On success, the number of events. On failure, -2.
This macro provides access to the “value” of the event. The semantics of the value varies according to the type of the event, and is described in the section on event codes (Event codes).
Next: Symbol nulled events, Previous: Basic event accessors, Up: Events [Contents][Index]
Allows the user to deactivate and reactivate symbol completion events in the grammar. When a recognizer is created, the activation status of each of its events is initialized to the activation status of that event in the base grammar. If reactivate is zero, the event is deactivated in the grammar. If reactivate is one, the event is activated in the grammar.
Symbol completion events are active by default if the symbol was set up for completion events in the grammar. If a symbol was not set up for completion events in the grammar, symbol completion events are inactive by default and any attempt to change that is a fatal error.
The activation status of a completion event in the grammar can only be changed if the symbol is marked as a completion event symbol in the grammar, and before the grammar is precomputed. However, if a symbol is marked as a completion event symbol in the recognizer, the completion event can be deactivated and reactivated in the recognizer.
Success cases: On success, the method returns the value of reactivate. The method succeeds trivially if the symbol is already set as indicated by reactivate.
Failure cases: If the active status of the completion event for sym_id cannot be set as indicated by reactivate, the method fails. On failure, -2 is returned.
Allows the user to deactivate and reactivate symbol completion events in the recognizer. If reactivate is zero, the event is deactivated. If reactivate is one, the event is activated.
Symbol completion events are active by default if the symbol was set up for completion events in the grammar. If a symbol was not set up for completion events in the grammar, symbol completion events are inactive by default and any attempt to change that is a fatal error.
Success cases: On success, the method returns the value of reactivate. The method succeeds trivially if the symbol is already set as indicated by reactivate.
Failure cases: If the active status of the completion event for sym_id cannot be set as indicated by reactivate, the method fails. On failure, -2 is returned.
Libmarpa can be set up to generate an
MARPA_EVENT_SYMBOL_COMPLETED
event whenever the symbol is completed.
A symbol is said to be completed
when a non-nulling rule with
that symbol on its LHS is completed.
For completion events to occur, the symbol must be marked
as a completion event symbol.
The
marpa_g_symbol_is_completion_event_set()
function
marks symbol sym_id as a completion event symbol
if value is 1,
and unmarks it
it as a completion event symbol if value is 0.
The
marpa_g_symbol_is_completion_event()
method
returns the current value of the completion event marking
for symbol sym_id.
Marking a completion event sets its activation status to on. Unmarking a completion event sets its activation status to off. The completion event marking cannot be changed once the grammar is precomputed.
If a completion event is marked,
its activation status can be changed using the
marpa_g_completion_symbol_activate()
method.
Note that, if a symbol is marked as a completion event symbol
in the recognizer,
its completion event can be deactivated
and reactivated in the recognizer.
Nulled rules and symbols will never cause completion events. Nullable symbols may be marked as completion event symbols, but this will have an effect only when the symbol is not nulled. Nulling symbols may be marked as completion event symbols, but no completion events will ever be generated for a nulling symbol. Note that this implies at no completion event will ever be generated at earleme 0, the start of parsing.
Success: On success, 1 if symbol sym_id is a completion event symbol after the call, 0 otherwise.
Failures: If sym_id is well-formed, but there is no such symbol, -1. If the grammar g is precomputed; or on other failure, -2.
Next: Prediction events, Previous: Completion events, Up: Events [Contents][Index]
Allows the user to deactivate and reactivate symbol nulled events in the grammar. When a recognizer is created, the activation status of each of its events is initialized to the activation status of that event in the base grammar. If reactivate is zero, the event is deactivated in the grammar. If reactivate is one, the event is activated in the grammar.
Symbol nulled events are active by default if the symbol was set up for nulled events in the grammar. If a symbol was not set up for nulled events in the grammar, symbol nulled events are inactive by default and any attempt to change that is a fatal error.
The activation status of a nulled event in the grammar can only be changed if the symbol is marked as a nulled event symbol in the grammar, and before the grammar is precomputed. However, if a symbol is marked as a nulled event symbol in the recognizer, the nulled event can be deactivated and reactivated in the recognizer.
Success cases: On success, the method returns the value of reactivate. The method succeeds trivially if the symbol is already set as indicated by reactivate.
Failure cases: If the active status of the nulled event for sym_id cannot be set as indicated by reactivate, the method fails. On failure, -2 is returned.
Allows the user to deactivate and reactivate symbol nulled events in the recognizer. If boolean is zero, the event is deactivated. If boolean is one, the event is activated.
Symbol nulled events are active by default if the symbol was set up for nulled events in the grammar. If a symbol was not set up for nulled events in the grammar, symbol nulled events are inactive by default and any attempt to change that is a fatal error.
Success cases: On success, the method returns the value of boolean. The method succeeds trivially if the symbol is already set as indicated by boolean.
Failure cases: If the active status of the nulled event for sym_id cannot be set as indicated by boolean, the method fails. On failure, -2 is returned.
Libmarpa can set up to generate
an MARPA_EVENT_SYMBOL_NULLED
event whenever the symbol is nulled.
A symbol is said to be nulled
when a zero length instance of that symbol
is recognized.
For nulled events to occur, the symbol must be marked
as a nulled event symbol.
The
marpa_g_symbol_is_nulled_event_set()
function
marks symbol sym_id as a nulled event symbol
if value is 1,
and unmarks it
it as a nulled event symbol if value is 0.
The
marpa_g_symbol_is_nulled_event()
method
returns the current value of the nulled event marking
for symbol sym_id.
Marking a nulled event sets its activation status to on. Unmarking a nulled event sets its activation status to off. The nulled event marking cannot be changed once the grammar is precomputed.
If a nulled event is marked,
its activation status can be changed using the
marpa_g_nulled_symbol_activate()
method.
Note that, if a symbol is marked as a nulled event symbol
in the recognizer,
its nulled event can be deactivated
and reactivated in the recognizer.
As a reminder, a symbol instance is a symbol at a specific location in the input, and with a specific length. Also, whenever a nulled symbol instance is recognized at a location, it is acceptable at that location, and vice versa.
When a symbol instance is recognized at a location, it will generate a nulled event or a prediction event, but never both. A symbol instance of zero length, when recognized at a location, generates a nulled event at that location, and does not generate a completion event. A symbol instance of non-zero length, when acceptable at a location, generates a completion event at that location, and does not generate a nulled event.
When a symbol instance is acceptable at a location, it will generate a nulled event or a prediction event, but never both. A symbol instance of zero length, when acceptable at a location, generates a nulled event at that location, and does not generate a prediction event. A symbol instance of non-zero length, when acceptable at a location, generates a prediction event at that location, and does not generate a nulled event.
While it is not possible for a symbol instance to generate both a nulled event and a completion event at a location, it is quite possible that a symbol might generate both kinds of event at that location. This is because multiple instances of the same symbol may be recognized at a given location, and these instances will have different lengths. If one instance is recognized at a given location as zero length and a second, non-zero-length, instance is recognized at the same location, the first will generate only nulled events, while the second will generate only completion events. For similar reasons, while a symbol instance will never generate both a null event and a prediction event at a location, multiple instances of the same symbol may do so.
Zero length derivations can be ambiguous. When a zero length symbol is recognized, all of its zero-length derivations are also considered to be recognized.
The
marpa_g_symbol_is_nulled_event_set()
method will
mark a symbol as a nulled event symbol,
even if the symbol is non-nullable.
This is convenient, for example,
for automatically generated grammars.
Applications which wish to treat
it as a failure if there is an
attempt to
mark a non-nullable symbol
as a nulled event symbol,
can check for this case using
the
marpa_g_symbol_is_nullable()
method.
Success: On success, 1 if symbol sym_id is a nulled event symbol after the call, 0 otherwise.
Failures: If sym_id is well-formed, but there is no such symbol, -1. If the grammar g is precomputed; or on other failure, -2.
Next: Symbol expected events, Previous: Symbol nulled events, Up: Events [Contents][Index]
Allows the user to deactivate and reactivate symbol prediction events in the grammar. When a recognizer is created, the activation status of each of its events is initialized to the activation status of that event in the base grammar. If reactivate is zero, the event is deactivated in the grammar. If reactivate is one, the event is activated in the grammar.
Symbol prediction events are active by default if the symbol was set up for prediction events in the grammar. If a symbol was not set up for prediction events in the grammar, symbol prediction events are inactive by default and any attempt to change that is a fatal error.
The activation status of a prediction event in the grammar can only be changed if the symbol is marked as a prediction event symbol in the grammar, and before the grammar is precomputed. However, if a symbol is marked as a prediction event symbol in the recognizer, the prediction event can be deactivated and reactivated in the recognizer.
Success cases: On success, the method returns the value of reactivate. The method succeeds trivially if the symbol is already set as indicated by reactivate.
Failure cases: If the active status of the prediction event for sym_id cannot be set as indicated by reactivate, the method fails. On failure, -2 is returned.
Allows the user to deactivate and reactivate symbol prediction events in the recognizer. If boolean is zero, the event is deactivated. If boolean is one, the event is activated.
Symbol prediction events are active by default if the symbol was set up for prediction events in the grammar. If a symbol was not set up for prediction events in the grammar, symbol prediction events are inactive by default and any attempt to change that is a fatal error.
Success cases: On success, the method returns the value of boolean. The method succeeds trivially if the symbol is already set as indicated by boolean.
Failure cases: If the active status of the prediction event for sym_id cannot be set as indicated by boolean, the method fails. On failure, -2 is returned.
Libmarpa can be set up
to generate a
MARPA_EVENT_SYMBOL_PREDICTED
event when a non-nulled symbol is predicted.
A non-nulled symbol is said to be predicted
when a instance of it
is acceptable at the current
earleme according to the grammar.
Nulled symbols do not generate predictions.
For predicted events to occur, the symbol must be marked
as a predicted event symbol.
The
marpa_g_symbol_is_predicted_event_set()
function
marks symbol sym_id as a predicted event symbol
if value is 1,
and unmarks it
it as a predicted event symbol if value is 0.
The
marpa_g_symbol_is_predicted_event()
method
returns the current value of the predicted event marking
for symbol sym_id.
Marking a prediction event sets its activation status to on. Unmarking a prediction event sets its activation status to off. The prediction event marking cannot be changed once the grammar is precomputed.
If a prediction event is marked,
its activation status can be changed using the
marpa_g_prediction_symbol_activate()
method.
Note that, if a symbol is marked as a prediction event symbol
in the recognizer,
its prediction event can be deactivated
and reactivated in the recognizer.
Success: On success, 1 if symbol sym_id is a predicted event symbol after the call, 0 otherwise.
Failures: If sym_id is well-formed, but there is no such symbol, -1. If the grammar g is precomputed; or on other failure, -2.
Next: Event codes, Previous: Prediction events, Up: Events [Contents][Index]
Sets the “expected symbol event bit” for symbol_id to value. A recognizer event is created whenever symbol symbol_id is expected at the current earleme. if and only if the expected symbol event bit for symbol_id is 1. The “expected symbol event bit” must be 1 or 0.
In this context, “expected” means “expected as a terminal”. Even if a symbol is predicted at the current earleme, if it is not acceptable as a terminal, it does not trigger an “expected symbol event”.
By default, the “expected symbol event bit” is 0. It is an error to attempt to set the “expected symbol event bit” to 1 for a nulling symbol, an inaccessible symbol, or an unproductive symbol.
Return value: The value of the event bit after the method call is finished. -2 if symbol_id is not the ID of a valid symbol; if it is the ID of an nulling, inaccessible for unproductive symbol; or on other failure.
Previous: Symbol expected events, Up: Events [Contents][Index]
Applications should never see this event. Event value: Undefined. Suggested message: "No event".
A nullable symbol is either the separator for, or the right hand side of, a sequence. Event value: The ID of the symbol. Suggested message: "This symbol is a counted nullable".
This event indicates that an application-settable threshold on the number of Earley items has been reached or exceeded. See marpa_r_earley_item_warning_threshold_set().
Event value: The current Earley item count. Suggested message: "Too many Earley items".
The parse is exhausted. Event value: Undefined. Suggested message: "Recognizer is exhausted".
One or more rules are loop rules — rules that are part of a cycle. Cycles are pathological cases of recursion, in which the same symbol string derives itself a potentially infinite number of times. Nonetheless, Marpa parses in the presence of these, and it is up to the application to treat these as fatal errors, something they almost always will wish to do. Event value: The count of loop rules. Suggested message: "Grammar contains a infinite loop".
A nulling symbol is also a terminal. Event value: The ID of the symbol. Suggested message: "This symbol is a nulling terminal".
The recognizer can be set to generate an event
a symbol is completed
using its
marpa_g_symbol_is_completion_event_set()
method.
(A symbol is "completed" if and only if any rule with that symbol
as its LHS is completed.)
This event code indicates that one of those events
occurred.
Event value: The ID of the completed symbol.
Suggested message: "Completed symbol".
The recognizer can be set to generate an event when a
symbol is expected as a terminal,
using its
marpa_r_expected_symbol_event_set()
method.
Note that this event only triggers if the symbol is
expected as a terminal.
Predicted symbols which are not expected as terminals
do not trigger this event.
This event code indicates that one of those events
occurred.
Event value: The ID of the expected symbol.
Suggested message: "Expecting symbol".
The recognizer can be set to generate an event when a
symbol is nulled – that is, recognized as a
zero-length symbol.
To set an nulled symbol event,
use the recognizer’s
marpa_r_nulled_symbol_event_set()
method.
This event code indicates that a nulled symbol event
occurred.
Event value: The ID of the nulled symbol.
Suggested message: "Symbol was nulled".
The recognizer can be set to generate an event when a
symbol is predicted.
To set an predicted symbol event,
use the recognizer’s
marpa_g_symbol_is_prediction_event_set()
method.
Unlike the
MARPA_EVENT_SYMBOL_EXPECTED
event,
the MARPA_EVENT_SYMBOL_PREDICTED
event
triggers for predictions of both
non-terminals and terminals.
This event code indicates that a predicted symbol event
occurred.
Event value: The ID of the predicted symbol.
Suggested message: "Symbol was predicted".
Next: Technical notes, Previous: Events, Up: Top [Contents][Index]
• Error methods: | ||
• Error Macros: | ||
• External error codes: | ||
• Internal error codes: |
Next: Error Macros, Previous: Error methods macros and codes, Up: Error methods macros and codes [Contents][Index]
When a method fails,
this method allows the application to read
the error code.
p_error_string is reserved for use by
the internals.
Applications should set it to NULL
.
Return value: The last error code from a Libmarpa method. Always succeeds.
Sets the error code
to MARPA_ERR_NONE
.
Not often used,
but now and then it can be useful
to force the error code to a known state.
Return value: MARPA_ERR_NONE
.
Always succeeds.
Next: External error codes, Previous: Error methods, Up: Error methods macros and codes [Contents][Index]
The number of error codes.
All error codes, whether internal or external,
will be integers, non-negative but
strictly less than MARPA_ERRCODE_COUNT
.
Next: Internal error codes, Previous: Error Macros, Up: Error methods macros and codes [Contents][Index]
This section lists the external error codes. These are the only error codes that users of the Libmarpa external interface should ever see. Internal error codes are in their own section (Internal error codes).
No error condition.
The error code is initialized to this value.
Methods which do not result in failure
sometimes reset the error code to MARPA_ERR_NONE
.
Numeric value: 0.
Suggested message: "No error".
A separator was specified for a sequence rule, but its ID was not that of a valid symbol. Numeric value: 6. Suggested message: "Separator has invalid symbol ID".
A tree iterator is positioned before the first tree,
and it was specified in a context where that is not
allowed.
A newly created tree is positioned before the first
tree.
To position a newly created tree iterator to the first tree
use the
marpa_t_next()
method.
Numeric value: 91.
Suggested message: "Tree iterator is before first tree".
A “counted” symbol was found that is also a nullable symbol. A “counted” symbol is one that appears on the RHS of a sequence rule. If a symbol is nullable, counting its occurrences becomes difficult. Questions of definition and problems of implementation arise. At a minimum, a sequence with counted nullables would be wildly ambigious.
Sequence rules are simply an optimized shorthand for rules that can also be written in ordinary BNF. If the equivalent of a sequence of nullables is really what your application needs, nothing in Libmarpa prevents you from specifying that sequence with ordinary BNF rules.
Numeric value: 8. Suggested message: "Nullable symbol on RHS of a sequence rule".
This error indicates an attempt to add a BNF rule which is a duplicate of a BNF rule already in the grammar. Two BNF rules are considered duplicates if
Duplication of sequence rules, and duplication between BNF rules and sequence rules, is dealt with by requiring that the LHS of a sequence rule not be the LHS of any other rule.
Numeric value: 11. Suggested message: "Duplicate rule".
This error indicates an attempt to add a duplicate token. A token is a duplicate if one already read at the same earleme has the same symbol ID and the same length. Numeric value: 12. Suggested message: "Duplicate token".
This error code indicates that an implementation-defined limit on the number of Earley items per Earley set was exceedeed. This limit is different from the Earley item warning threshold, an optional limit on the number of Earley items in an Earley set, which can be set by the application.
The implementation defined-limit is very large, at least 500,000,000 earlemes. An application is unlikely ever to see this error. Libmarpa’s use of memory would almost certainly exceed the implementation’s limits before it occurred. Numeric value: 13. Suggested message: "Maximum number of Earley items exceeded".
A negative event index was specified. That is not allowed. Numeric value: 15. Suggested message: "Negative event index".
An non-negative event index was specified, but there is no event at that index. Since the events are in sequence, this means it was too large. Numeric value: 16. Suggested message: "No event at that index".
The grammar has a cycle — one or more loop rules. This is a recoverable error, although most applications will want to treat it as fatal. For more see the description of marpa_g_precompute. Numeric value: 17. Suggested message: "Grammar has cycle".
This is an internal error, and indicates that Libmarpa was wrongly built. Libmarpa was compiled with headers which do not match the rest of the code. The solution is to find a correctly built Libmarpa. Numeric value: 98. Suggested message: "Internal error: Libmarpa was built incorrectly"
The Libmarpa base grammar is in a "not ok" state. Currently, the only way this can happen is if Libmarpa memory is being overwritten. Numeric value: 29. Suggested message: "Marpa is in a not OK state".
This error code indicates that the token symbol is an inaccessible symbol — one which cannot be reached from the start symbol.
Since the inaccessibility of a symbol is a property of the grammar, this error code typically indicates an application error. Nevertheless, a retry at this location, using another token ID, may succeed. At this writing, the author knows of no uses of this technique.
Numeric value: 18. Suggested message: "Token symbol is inaccessible".
A function was called that takes a boolean argument, but the value of that argument was not either 0 or 1. Numeric value: 22. Suggested message: "Argument is not boolean".
The location (Earley set ID) is not valid. It may be invalid for one of two reasons:
For users of input models other than the standard one, the term “location”, as used in association with this error code, means Earley set ID or Earley set ordinal. In the standard input model, this will always be identical with Libmarpa’s other idea of location, the earleme.
Numeric value: 25. Suggested message: "Location is not valid".
A start symbol was specified, but its symbol ID is not that of a valid symbol. Numeric value: 27. Suggested message: "Specified start symbol is not valid".
A method was called with an invalid assertion ID. This is a assertion ID which not only does not exist, but cannot exist. Currently that means its value is less than zero. Numeric value: 96. Suggested message: "Assertion ID is malformed".
A method was called with an invalid rule ID. This is a rule ID which not only does not exist, but cannot exist. Currently that means its value is less than zero. Numeric value: 26. Suggested message: "Rule ID is malformed".
A method was called with an invalid symbol ID. This is a symbol ID which not only does not exist, but cannot exist. Currently that means its value is less than zero. Numeric value: 28. Suggested message: "Symbol ID is malformed".
There was a mismatch in the major version number between the requested version of libmarpa, and the actual one. Numeric value: 30. Suggested message: "Libmarpa major version number is a mismatch".
There was a mismatch in the micro version number between the requested version of libmarpa, and the actual one. Numeric value: 31. Suggested message: "Libmarpa micro version number is a mismatch".
There was a mismatch in the minor version number between the requested version of libmarpa, and the actual one. Numeric value: 32. Suggested message: "Libmarpa minor version number is a mismatch".
A non-negative Earley set ID (also called an Earley set ordinal) was specified, but there is no corresponding Earley set. Since the Earley set ordinals are in sequence, this means that the specified ID is greater than that of the latest Earley set. Numeric value: 39. Suggested message: "Earley set ID is after latest Earley set".
The grammar is not precomputed, and attempt was made to do something with it that is not allowed for unprecomputed grammars. For example, a recognizer cannot be created from a grammar until it is precomputed. Numeric value: 34. Suggested message: "This grammar is not precomputed".
The application attempted to create a bocage from a recognizer without a parse. Applications will often want to treat this as a soft error. Numeric value: 41. Suggested message: "No parse".
A grammar which has no rules is being used in a way that is not allowed. Usually the problem is that the user is trying to precompute the grammar. Numeric value: 42. Suggested message: "This grammar does not have any rules".
The grammar has no start symbol, and an attempt was made to perform an operation which requires one. Usually the problem is that the user is trying to precompute the grammar. Numeric value: 43. Suggested message: "This grammar has no start symbol".
A method was called with an assertion ID which is well-formed, but the assertion does not exist. Numeric value: 97. Suggested message: "No assertion with this ID exists".
A method was called with a rule ID which is well-formed, but the rule does not exist. Numeric value: 89. Suggested message: "No rule with this ID exists".
A method was called with a symbol ID which is well-formed, but the symbol does not exist. Numeric value: 90. Suggested message: "No symbol with this ID exists".
This error code indicates that no tokens at all were expected at this earleme location. This can only happen in alternative input models.
Typically, this indicates an application programming error. Retrying input at this location will always fail. But if the application is able to leave this earleme empty, a retry at a later location, using this or another token, may succeed. At this writing, the author knows of no uses of this technique.
Numeric value: 44. Suggested message: "No token is expected at this earleme location".
This error occurs in situations where a rule is required to be a sequence, and indicates that the rule of interest is, in fact, not a sequence.
Numeric value: 99. Suggested message: "Rule is not a sequence".
Marpa does not allow a symbol to be both nulling and a terminal. Numeric value: 49. Suggested message: "A symbol is both terminal and nulling".
The Marpa order object has been frozen. If a Marpa order object is frozen, it cannot be changed.
Multiple tree iterators can share a Marpa order object, but that order object is frozen after the first tree iterator is created from it. Applications can order an bocage in many ways, but they must do so by creating multiple order objects.
Numeric value: 50. Suggested message: "The ordering is frozen".
The parse is exhausted. Numeric value: 53. Suggested message: "The parse is exhausted".
The parse is too long. The limit on the length of a parse is implementation dependent, but it is very large, at least 500,000,000 earlemes.
This error code is unlikely in the standard input model. Almost certainly memory would be exceeded before it could occur. If an application sees this error, it almost certainly using one of the non-standard input models.
Most often this message will occur because of a request to add a single extremely long token, perhaps as a result of an application error. But it is also possible this error condition will occur after the input of a large number of long tokens.
Numeric value: 54. Suggested message: "This input would make the parse too long".
In a method which takes pointers as arguments,
one of the pointer arguments is NULL
,
in a case where that is not allowed.
One such method is marpa_r_progress_item()
.
Numeric value: 56.
Suggested message: "An argument is null when it should not be".
An attempt was made to use a precomputed grammar in a way that is not allowed. Often this is an attempt to change the grammar. Nearly every change to a grammar after precomputation invalidates the precomputation, and is therefore not allowed. Numeric value: 57. Suggested message: "This grammar is precomputed".
No recognizer progress report is currently active,
and an action has been attempted which
is inconsistent with that.
One such action would be a
marpa_r_progress_item()
call.
Numeric value: 59.
Suggested message: "No progress report has been started".
The progress report is “exhausted” — all its items have been iterated through. Numeric value: 58. Suggested message: "The progress report is exhausted".
A symbol or rule rank was specified which was less than an implementation-defined minimum. Implementations will always allow at least those ranks in the range between -134,217,727 and 134,217,727. Numeric value: 85. Suggested message: "Rule or symbol rank too low".
A symbol or rule rank was specified which was greater than an implementation-defined maximum. Implementations will always allow at least those ranks in the range between -134,217,727 and 134,217,727. Numeric value: 86. Suggested message: "Rule or symbol rank too high".
The recognizer is “inconsistent”,
usually because the user has rejected one or
more rules or terminals,
and has not yet called
the
marpa_r_consistent()
method.
Numeric value: 95.
Suggested message: "The recognizer is inconsistent.
The recognizer is not accepting input, and the application has attempted something that is inconsistent with that fact. Numeric value: 60. Suggested message: "The recognizer is not accepting input".
The recognizer has not been started. and the application has attempted something that is inconsistent with that fact. Numeric value: 61. Suggested message: "The recognizer has not been started".
The recognizer has been started. and the application has attempted something that is inconsistent with that fact. Numeric value: 62. Suggested message: "The recognizer has been started".
The index of a RHS symbol was specified, and it was negative. That is not allowed. Numeric value: 63. Suggested message: "RHS index cannot be negative".
A non-negative index of RHS symbol was specified, but there is no symbol at that index. Since the indexes are in sequence, this means the index was greater than or equal to the rule length. Numeric value: 64. Suggested message: "RHS index must be less than rule length".
An attempt was made to add a rule with too many right hand side symbols. The limit on the RHS symbol count is implementation dependent, but it is very large, at least 500,000,000 symbols. This is far beyond what is required in any current practical grammar. An application with rules of this length is almost certain to run into memory and other limits. Numeric value: 65. Suggested message: "The RHS is too long".
The LHS of a sequence rule cannot be the LHS of any other rule, whether a sequence rule or a BNF rule. An attempt was made to violate this restriction. Numeric value: 66. Suggested message: "LHS of sequence rule would not be unique".
The start symbol is not on the LHS on any rule. That means it could never match any possible input, not even the null string. Presumably, an error in writing the grammar. Numeric value: 73. Suggested message: "Start symbol not on LHS of any rule".
An attempt was made to use a symbol in a way that requires it to be set up for completion events, but the symbol was not set set up for completion events. The archetypal case is an attempt to activate completion events for the symbol in the recognizer. The archetypal case is an attempt to activate a completion event in the recognizer for a symbol that is not set up as a completion event. Numeric value: 92. Suggested message: "Symbol is not set up for completion events".
An attempt was made to use a symbol in a way that requires it to be set up for nulled events, but the symbol was not set set up for nulled events. The archetypal case is an attempt to activate a nulled events in the recognizer for a symbol that is not set up as a nulled event. Numeric value: 93. Suggested message: "Symbol is not set up for nulled events".
An attempt was made to use a symbol in a way that requires it to be set up for predictino events, but the symbol was not set set up for predictino events. The archetypal case is an attempt to activate a prediction event in the recognizer for a symbol that is not set up as a prediction event. Numeric value: 94. Suggested message: "Symbol is not set up for prediction events".
Unvalued symbols are a deprecated Marpa feature,
which may be avoided with
the
marpa_g_force_valued()
method.
An unvalued symbol may take on any value,
and therefore a symbol which is unvalued at some points
cannot safely to be used to contain a value at
others.
This error indicates that such an unsafe use is
being attempted.
Numeric value: 74.
Suggested message: "Symbol is treated both as valued and unvalued".
An attempt was made to change the terminal status of a symbol to a different value after it was locked. Numeric value: 75. Suggested message: "The terminal status of the symbol is locked".
A token was specified whose symbol ID is not a terminal. Numeric value: 76. Suggested message: "Token symbol must be a terminal".
A token length was specified which is less than or equal to zero. Zero-length tokens are not allowed in Libmarpa. Numeric value: 77. Suggested message: "Token length must greater than zero".
The token length is too long. The limit on the length of a token is implementation dependent, but it is at least 500,000,000 earlemes. An application using a token that long is almost certain to run into some other limit. Numeric value: 78. Suggested message: "Token is too long".
A Libmarpa parse tree iterator is “exhausted”, that is, it has no more parses. Numeric value: 79. Suggested message: "Tree iterator is exhausted".
A Libmarpa tree is “paused”
and an operation was attempted which
is inconsistent with that fact.
Typically, this operation will be
a call of the
marpa_t_next()
method.
Numeric value: 80.
Suggested message: "Tree iterator is paused".
An attempt was made to read a token where a token with that symbol ID is not expected. This message can also occur when an attempt is made to read a token at a location where no token is expected. Numeric value: 81. Suggested message: "Unexpected token".
The start symbol is unproductive. That means it could never match any possible input, not even the null string. Presumably, an error in writing the grammar. Numeric value: 82. Suggested message: "Unproductive start symbol".
The valuator is inactive in a context where that should not be the case. Numeric value: 83. Suggested message: "Valuator inactive".
Unvalued symbols are a deprecated Marpa feature,
which may be avoided with
the
marpa_g_force_valued()
method.
This error code
indicates that the valued status of a symbol is locked,
and an attempt was made
to change it to a status different from the
current one.
Numeric value: 84.
Suggested message: "The valued status of the symbol is locked".
An attempt was made to do something with a nulling
symbol that is not allowed.
For example,
the ID of a nulling symbol cannot be an argument
to
marpa_r_expected_symbol_event_set()
—
because it is not possible to create an “expected symbol” event
for a nulling symbol.
Numeric value: 87.
Suggested message: "Symbol is nulling".
An attempt was made to do something with an unused symbol that is not allowed.
An “unused” symbol is a inaccessible or unproductive symbol.
For example,
the ID of a unused symbol cannot be an argument
to
marpa_r_expected_symbol_event_set()
—
because it is not possible to create an “expected symbol” event
for an unused symbol.
Numeric value: 88.
Suggested message: "Symbol is not used".
Previous: External error codes, Up: Error methods macros and codes [Contents][Index]
An internal error code may be one of two things: First, it can be an error code which arises from an internal Libmarpa programming issue (in other words, something happening in the code that was not supposed to be able to happen.) Second, it can be an error code which only occurs when a method from Libmarpa’s internal interface is used. Both kinds of internal error message share one common trait — users of the Libmarpa’s external interface should never see them.
Internal error messages require someone with knowledge of the Libmarpa internals to follow up on them. They usually do not have descriptions or suggested messages.
Numeric value: 1.
Numeric value: 2.
Numeric value: 3.
Numeric value: 4.
Numeric value: 5.
Numeric value: 7.
"Development" errors were used heavily during Libmarpa’s development, when it was not yet clear how precisely to classify every error condition. Unless they are using a developer’s version, users of the external interface should never see development errors.
Development errors have an error string associated with them. The error string is a short 7-bit ASCII error string which describes the error. Numeric value: 9. Suggested message: "Development error, see string".
Numeric value: 10.
Numeric value: 14.
A “catchall” internal error. Numeric value: 19.
The AHFA ID was invalid. There are no AHFAs any more, so this message should not occur. Numeric value: 20.
The AHM ID was invalid. The term “AIMID” is a legacy of earlier implementations and must be kept for backward compatibility. Numeric value: 21.
Numeric value: 23.
Numeric value: 24.
Numeric value: 33.
Numeric value: 35.
Numeric value: 36.
Numeric value: 37.
Numeric value: 38.
Numeric value: 40.
Numeric value: 46.
Numeric value: 47.
Numeric value: 45.
Numeric value: 48.
Numeric value: 51.
Numeric value: 52.
Numeric value: 55.
Numeric value: 70.
Numeric value: 71.
Numeric value: 68.
Numeric value: 69.
Numeric value: 67.
Numeric value: 72.
Next: Advanced input models, Previous: Error methods macros and codes, Up: Top [Contents][Index]
This section contains technical notes that are not necessary for the main presentation, but which may be useful or interesting.
• Data types used by Libmarpa: | ||
• Why so many time objects: | ||
• Design of numbered objects: | ||
• LHS Terminals: |
Next: Why so many time objects, Previous: Technical notes, Up: Technical notes [Contents][Index]
Libmarpa does not use any floating point data or strings. All data are either integers or pointers.
Next: Design of numbered objects, Previous: Data types used by Libmarpa, Up: Technical notes [Contents][Index]
Marpa is an aggressively multi-pass algorithm. Marpa achieves its efficiency, not in spite of making multiple passes over the data, but because of it. Marpa regularly substitutes two fast O(n) passes for a single O(n log n) pass. Marpa’s proliferation of time objects is in keeping with its multi-pass approach.
Bocage objects come at no cost, even for unambiguous parses, because the same pass which creates the bocage also deals with other issues which are of major significance for unambiguous parses. It is the post-processing of the bocage pass that enables Marpa to do both left- and right-recursion in linear time.
Of the various objects, the best case for elimination is of the ordering object. In many cases, the ordering is trivial. Either the parse is unambiguous, or the application does not care about the order in which parses are returned. But while it would be easy to add an option to bypass creation of an ordering object, there is little to be gained from it. When the ordering is trivial, its overhead is very small — essentially a handful of subroutine calls. Many orderings accomplish nothing, but these cost next to nothing.
Tree objects come at minimal cost to unambiguous grammars, because the same pass that allows iteration through multiple parse trees does the tree traversal. This eliminates much of the work that otherwise would need to be done in the valuation time object. In the current implement, the valuation time object needs only to step through a sequence already determined in the tree iterator.
Next: LHS Terminals, Previous: Why so many time objects, Up: Technical notes [Contents][Index]
As the name suggests, the choice was made to implement numbered objects as integers, and not as pointers. In standard-conformant C, integers can be safely checked for validity, while pointers cannot.
There are efficiency tradeoffs between pointers and integers but they are complicated, and they go both ways. Pointers can be faster, but integers can be used as indexes into more than one data structure. Which is actually faster depends on the design. Integers allow for a more flexible design, so that once the choice is settled on, careful programming can make them a win, possibly a very big one.
The approach taken in Libmarpa was to settle, from the outset, on integers as the implementation for numbered objects, and to optimize on that basis. The author concedes that it is possible that others redoing Libmarpa from scratch might find that pointers are faster. But the author is confident that they will also discover, on modern architectures, that the lack of safe validity checking is far too high a price to pay for the difference in speed.
Previous: Design of numbered objects, Up: Technical notes [Contents][Index]
Marpa’s idea in losing the sharp division between terminals and non-terminals is that the distinction, while helpful for proving theorems, is not essential in practice. LHS symbols in the input might be useful for “short circuiting” the rules in which they occur. This may prove helpful in debugging, or have other applications.
However, it also can be useful, for checking input validity as well as for efficiency, to follow tradition and distinguish non-terminals from terminals. For this reason, the traditional behavior is the default in Libmarpa.
Next: Futures, Previous: Technical notes, Up: Top [Contents][Index]
In an earlier chapter, we introduced Libmarpa’s concept of input, and described its basic input models. See Input. In this chapter we describe Libmarpa’s advanced models of input. These advanced input models have attracted considerable interest. However, they have seen little actual use so far, and for that reason we delayed their consideration until now.
A Libmarpa input model is advanced if it allows tokens of length other than 1. The advanced input models are also called variable-length token models because they allow the token length to vary from the “normal” length of 1.
• The dense variable-length token model: | ||
• The fully general input model: |
Next: The fully general input model, Previous: Advanced input models, Up: Advanced input models [Contents][Index]
In the
dense variable-length model of input,
one or more successful
calls of
marpa_r_alternative()
must be immediately previous
to every call to
marpa_r_earleme_complete()
.
Note that,
for a variable-length input model to be “dense”
according to this definition,
at least one successful call
of marpa_r_alternative()
must be immediately previous to each call to
marpa_r_earleme_complete()
.
Recall that, in this document, we say that a marpa_r_alternative()
call is
“immediately previous” to a
marpa_r_earleme_complete()
call
iff
that marpa_r_earleme_complete()
call is
the first
marpa_r_earleme_complete()
call after
the marpa_r_alternative()
call.
In the dense model of input,
after a successful call of
marpa_r_alternative()
,
the earleme variables are as follows:
max(old_f, old_c+length)
,
marpa_r_alternative()
,
marpa_r_alternative()
, and
marpa_r_alternative()
never changes the
latest or current earleme.
In the dense variable-length model of input,
the effect of the
marpa_r_earleme_complete()
mutator on the earleme variables
is the same as for the
basic models of input.
See The standard model of input.
In the dense model of input, the latest earleme is always the same as the current earleme. In fact, the latest earleme and the current earleme are always the same, except in the fully general model of input.
Previous: The dense variable-length token model, Up: Advanced input models [Contents][Index]
In the
sparse variable-length model of input,
zero or more successful
calls of
marpa_r_alternative()
must be immediately previous
to every call to
marpa_r_earleme_complete()
.
The sparse model is the dense variable-length model,
with its only restriction lifted —
the sparse variable-length input model
allows calls to
marpa_r_earleme_complete()
that are not immediately preceded by calls to
marpa_r_alternative()
.
Since it is unrestricted, the sparse input model is Libmarpa’s fully general input model. Because of this, it may be useful for us specify the effect of mutators on the earleme variables in detail, even at the expense of some repetition.
In the sparse input model,
empty earlemes
are now possible.
An empty earleme is an earleme
with no tokens and no Earley set.
An empty earleme occurs iff
marpa_r_earleme_complete()
is called when there is no immediately previous
call to
marpa_r_alternative()
.
The sparse model takes its name
from the fact that there may be earlemes with no
Earley set.
In the sparse model, Earley sets are “sparsely”
distributed among the earlemes.
In the dense model of input,
the effect on the earleme variables of
a successful call of the
marpa_r_alternative()
mutator
is the same as for the sparse model of input:
max(old_f, old_c+length)
,
marpa_r_alternative()
,
marpa_r_alternative()
, and
marpa_r_alternative()
never changes the
latest or current earleme.
In the sparse model,
when the earleme is not empty,
the effect of
a call to
marpa_r_earleme_complete()
on the earleme variables is the same as
in the dense and the basic models of input.
Specifically, the following will be true:
old_c+1
,
where old_c is the current earleme before the call.
old_c+1
, and therefore
will be equal to the current earleme.
marpa_r_earleme_complete()
.
Recall that, in the dense and basic input models,
as a matter of definition,
there are no empty earlemes.
For the sparse input model,
in the case of an empty earleme,
the effect of the
marpa_r_earleme_complete()
mutator on the earleme variables
is the following:
old_c+1
,
where old_c is the current earleme before the call.
marpa_r_earleme_complete()
.
After a call to marpa_r_earleme_complete()
for an empty earleme,
the lastest and current earlemes will have different values.
In a parse that never calls marpa_r_earleme_complete()
for an empty earleme,
the lastest and current earlemes will always be the same.
Next: Deprecated techniques and methods, Previous: Advanced input models, Up: Top [Contents][Index]
This chapter discusses features that are not in the external interface, but that might be added to the external interface in the future.
• Orthogonal treatment of exhaustion: | ||
• Furthest earleme values: | ||
• Additional recoverable failures in marpa_r_alternative(): | ||
• Untested methods: |
Next: Furthest earleme values, Previous: Futures, Up: Futures [Contents][Index]
The treatment of parse exhaustion is very awkward.
marpa_r_start_input()
returns success on exhaustion,
while
marpa_r_earleme_complete()
either returns success or
a hard failure, depending on circumstances.
See marpa_r_earleme_complete() and
marpa_r_start_input().
Ideally the treatment should be simpler, more intuitive and more orthogonal. Better, perhaps, would be to always treat parse exhaustion as a soft failure.
Next: Additional recoverable failures in marpa_r_alternative(), Previous: Orthogonal treatment of exhaustion, Up: Futures [Contents][Index]
marpa_r_furthest_earleme
returns
unsigned int
which is non-orthogonal with
marpa_r_current_earleme
.
This leaves no room for an failure return value,
which we deal with by not checking for failures,
of which the only important one is calling
marpa_r_furthest_earleme
before the start of input.
To consider
marpa_r_furthest_earleme
we consider furthese earleme to have
been initialized when the recognizer was created,
which is another non-orthogonality with
marpa_r_current_earleme
.
All this might be fine, if something were gained, but in fact in the furthest
earleme, unless there is a problem, always becomes the current earleme,
and no use cases for extremely long variable-length tokens are envisioned,
so that the two should never be far apart.
Additionally, the additional values for the furthest earleme only come into
play if the parse is to large for the computer memories as of this writing.
Summarizing, marpa_r_furthest_earleme
,
should return an int
,
like marpa_r_current_earleme
,
and the non-orthogonalities should be eliminated.
Next: Untested methods, Previous: Furthest earleme values, Up: Futures [Contents][Index]
Among the hard failures that
marpa_r_alternative() returns
are the error codes
MARPA_ERR_DUPLICATE_TOKEN
,
MARPA_ERR_NO_TOKEN_EXPECTED_HERE
and MARPA_ERR_INACCESSIBLE_TOKEN
.
These are currently irrecoverable.
They may in fact be fully recoverable,
but are not documented as such because this has not been
tested.
At this writing, we know of no applications which attempt to recover from these errors. It is possible that these error codes may also be useable for the techniques similar to the Ruby Slippers, as of this writing, we know of no proposals to use them in this way.
Previous: Additional recoverable failures in marpa_r_alternative(), Up: Futures [Contents][Index]
The methods of this section are not in the external interface, because they have not been adequately tested. Their fate is uncertain. Users should regard these methods as unsupported.
• Ranking methods: | ||
• Zero-width assertion methods: | ||
• Methods for revising parses: |
Next: Zero-width assertion methods, Previous: Untested methods, Up: Untested methods [Contents][Index]
These methods, respectively, set and query the default rank of the grammar. When a grammar is created, the default rank is 0. When rules and symbols are created, their rank is the default rank of the grammar.
Changing the grammar’s default rank does not affect those rules and symbols already created, only those that will be created. This means that the grammar’s default rank can be used to, in effect, assign ranks to groups of rules and symbols. Applications may find this behavior useful.
Return value: On success, returns
the rank after
the call,
and sets the error code to
MARPA_ERR_NONE
.
On failure, returns -2,
and sets the error code to an appropriate
value, which will never be
MARPA_ERR_NONE
.
Note that when the rank is -2,
the error code is the only way to distinguish
success from failure.
The error code can be determined by using the
marpa_g_error()
call.
These methods, respectively, set and query the rank of a symbol sym_id. When sym_id is created, its rank initialized to the default rank of the grammar.
Return value: On success, returns
the rank after
the call,
and sets the error code to
MARPA_ERR_NONE
.
On failure, returns -2,
and sets the error code to an appropriate
value, which will never be
MARPA_ERR_NONE
.
Note that when the rank is -2,
the error code is the only way to distinguish
success from failure.
The error code can be determined by using the
marpa_g_error()
call.
Next: Methods for revising parses, Previous: Ranking methods, Up: Untested methods [Contents][Index]
On success, returns previous default value of the assertion.
Changes default value to default_value. On success, returns previous default value of the assertion.
Previous: Zero-width assertion methods, Up: Untested methods [Contents][Index]
Marpa allows an application to “change its mind” about a parse, rejected rule previously recognized or predicted, and terminals previously scanned. The methods in this section provide that capability.
Next: Index of terms, Previous: Futures, Up: Top [Contents][Index]
• Valued and unvalued symbols: |
Previous: Deprecated techniques and methods, Up: Deprecated techniques and methods [Contents][Index]
• What unvalued symbols were: | ||
• Grammar methods dealing with unvalued symbols: | ||
• Registering semantics in the valuator: |
Next: Grammar methods dealing with unvalued symbols, Previous: Valued and unvalued symbols, Up: Valued and unvalued symbols [Contents][Index]
Libmarpa symbols can have values, which is the traditional way of doing semantics. Libmarpa also allows symbols to be unvalued. An unvalued symbol is one whose value is unpredictable from instance to instance. If a symbol is unvalued, we sometimes say that it has “whatever” semantics.
Situations where the semantics can tolerate unvalued symbols are surprisingly frequent. For example, the top-level of many languages is a series of major units, all of whose semantics are typically accomplished via side effects. The compiler is typically indifferent to the actual value produced by these major units, and tracking them is a waste of time. Similarly, the value of the separators in a list is typically ignored.
Rules are unvalued if and only if their LHS symbols are unvalued. When rules and symbols are unvalued, Libmarpa optimizes their evaluation.
It is in principle unsafe to check the value of a symbol if it can be unvalued. For this reason, once a symbol has been treated as valued, Libmarpa marks it as valued. Similarly, once a symbol has been treated as unvalued, Libmarpa marks it as unvalued. Once marked, a symbol’s valued status is locked and cannot be changed later.
The valued status of terminals is marked the first time they are read. The valued status of LHS symbols must be explicitly marked by the application when initializing the valuator — this is Libmarpa’s equivalent of registering a callback.
LHS terminals are disabled by default. If allowed, the user should be aware that the valued status of a LHS terminal will be locked in the recognizer if it is used as a terminal, and the symbol’s use as a rule LHS in the valuator must be consistent with the recognizer’s marking.
Marpa reports an error when a symbol’s use conflicts with its locked valued status. Doing so usually saves the Libmarpa user some tricky debugging further down the road.
Next: Registering semantics in the valuator, Previous: What unvalued symbols were, Up: Valued and unvalued symbols [Contents][Index]
These methods, respectively, set
and query the “valued status” of a symbol.
Once set to a value with the
marpa_g_symbol_is_valued_set()
method,
the valued status of a symbol is “locked” at that value.
It cannot thereafter be changed.
Subsequent calls to
marpa_g_symbol_is_valued_set()
for the same sym_id will fail,
leaving sym_id’s valued status unchanged,
unless value is the same as the locked-in value.
Return value: On success, 1 if the symbol symbol_id is valued after the call, 0 if not. If the valued status is locked and value is different from the current status, -2. If value is not 0 or 1; or on other failure, -2.
Previous: Grammar methods dealing with unvalued symbols, Up: Valued and unvalued symbols [Contents][Index]
By default, Libmarpa’s valuator objects
assume that
non-terminal symbols have
no semantics.
The archetypal application will need to register
symbols that contain semantics.
The primary method for doing this is
marpa_v_symbol_is_valued()
.
Applications will typically register semantics by rule,
and these applications will find
the
marpa_v_rule_is_valued()
method more convenient.
These methods, respectively,
set and query
the valued status of symbol sym_id.
marpa_v_symbol_is_valued_set()
will set
the valued status to
the value of
its status argument.
A valued status of 1 indicates that the symbol is valued.
A valued status of 0 indicates that the symbol is unvalued.
If the valued status is locked,
an attempt to change to a status different from the
current one will fail
(error code MARPA_ERR_VALUED_IS_LOCKED
).
Return value: On success, the valued status after the call. If value is not either 0 or 1, or on other failure, -2.
These methods, respectively,
set and query
the valued status
for the LHS symbol of rule rule_id.
marpa_v_rule_is_valued_set()
sets
the valued status to the value
of its status argument.
A valued status of 1 indicates that the symbol is valued.
A valued status of 0 indicates that the symbol is unvalued.
If the valued status is locked,
an attempt to change to a status different from the
current one will fail
(error code MARPA_ERR_VALUED_IS_LOCKED
).
Rules have no valued status of their own. The valued status of a rule is always that of its LHS symbol. These methods are conveniences — they save the application the trouble of looking up the rule’s LHS.
Return value: On success, the valued status of the rule rule_id’s LHS symbol after the call. If value is not either 0 or 1, or on other failure, -2.
This methods locks the valued status of all symbols to 1, indicated that the symbol is valued. If this is not possible, for example because one of the grammar’s symbols already is locked at a valued status of 0, failure is returned.
Return value: On success, a non-negative number.
On failure, returns -2,
and sets the error code to an appropriate
value, which will never be
MARPA_ERR_NONE
.
Previous: Deprecated techniques and methods, Up: Top [Contents][Index]
This index is of terms that are used in a special sense in this document. Not every use of these terms is indexed — only those uses which are in some way defining.
Jump to: | A B C D E F H I L M N O P R S T U V W |
---|
Jump to: | A B C D E F H I L M N O P R S T U V W |
---|