- About this document
- Marpa has two different ideas of location
- What is an alternative input model?
- Token streams
- The furthest earleme
- The latest Earley set and latest earleme
- Alternative models and read()
- Ambiguous lexing
- Duplicate tokens
- Earlemes: the details
- The character-per-earleme model
- Other input models
- Copyright and License
Marpa::R2::Advanced::Models - Other input models
The alternative input models described in this document are an advanced technique. If you are starting out with Marpa, you probably want to ignore this document. If you are an experienced Marpa user, it is still safe to ignore this document, but you might find the possibilities it discusses interesting.
In the other Marpa documentation, we have spoken of "location", and assumed the standard input model. The locations actually in use by the methods described for the standard input model were Earley set ordinals. (An Earley set's ordinal is also its ID.)
Marpa actually has two different ideas of location -- Earley set ordinal and earleme. This is ignored in the other Marpa documents, and it can be, because they assume the standard input model. Use of the standard input model guarantees that earleme and Earley set ordinal will always be exactly the same.
This document introduces methods which make it possible (and in fact likely) that earleme and Earley set ordinal will differ. From here on, the reader will need to pay careful attention to the distinction.
An alternative input model is anything that is not the default, token-stream model. More helpfully, Marpa allows variable-length tokens and ambiguous tokens, and an alternative input model is any input model which
Allows a token whose length is not exactly 1, or
Allows locations which have more than one token.
To do either of these things, a user must use the recognizer's
alternative method. In other words, if you are not using the recognizer's
alternative method call, you are not using an alternative input method.
Many concepts, such as parsing location, parse exhaustion, and the end of parsing, are somewhat more complicated when alternative input models are involved. These concepts were explained in the main document for the recognizer on the assumption that the default input model was in use. This document revises those explanations as necessary to take into account the alternative input models.
Marpa's default input model is the traditional one -- a token stream. Token streams are very standard in parsing applications -- so much so that most texts do not take the trouble of defining the term. A token stream is input structured as a sequence of tokens, where each token occupies one location and every location has a token. In the token stream model, all tokens are of the same length.
Conventionally, all tokens are of length 1, and the token stream starts at location 0. Following this convention, the Nth token would start at location N-1 and end at location N. For example, the first token would start at location 0 and end at location 1. The second token would start at location 1 and end at location 2.
For most parsers, position is location in a token stream. To deal with variable-length and overlapping tokens, Marpa needs a more flexible idea of location.
Marpa's tracks position in terms of earlemes. Earlemes are named after Jay Earley, the inventor of the first algorithm in Marpa's lineage. Every token has a start earleme and an end earleme.
The token stream model may also be called the token-per-earleme model. In the token stream model, token location and earleme location are exactly identical an one-to-one basis. If a user's application uses the token stream model, he can ignore earlemes, and think entirely in terms of tokens and position in a token stream. Because of this, the main Marpa documents speak simply of the "location" in the parse.
The furthest earleme is the last earleme at which a token ends. In the default input model, the furthest earleme is not an important concept. In the default input model, the furthest earleme and the current earleme are always the same.
In alternative input models, tokens may be longer than 1 earleme, and the furthest earleme and the current earleme may be far apart. This becomes an issue when parsing is finished. Alternative input models use the recognizer's
end_input method to ensure that processing of input catches up to the furthest earleme.
The latest earleme is the earleme location of the latest Earley set. In the default input model, the latest earleme is always the same as the current earleme.
In alternative input models, there may not be an Earley set at a given earleme location. When that is the case for the current earleme, then the latest Earley set is not at the current earleme, and the latest earleme and current earlemes are different.
$recce->alternative( 'a', \42, 1 ) or return 'First alternative failed';
alternative method is the most general method for reading input, and is used in alternative input models. It takes three arguments, only the first of which is required.
The first two arguments are the token type and a reference to the token value. If the reference to the token value is omitted, or is
undef, the token has a "whatever" value.
The third argument to the
alternative method is a token length. If omitted, token length defaults to 1, which is the correct value for the token stream model. Its value can be any integer greater than zero. Marpa does not allow zero length tokens in any input model.
read method, the
alternative method does not advance the current location (current earleme) on each call. This allows the application to read several tokens at the same earleme. This is how ambiguous input is created. To advance the current earleme when input is read using the
alternative method, the
complete_earleme method must be called.
alternative returns a Perl true value. If the token is rejected,
alternative returns a Perl false. On other failures,
alternative throws an exception.
$current_earleme = $recce->current_earleme();
Returns the current parse location, also known as the current earleme. Not often needed.
my $origin_earleme = $recce->earleme($origin_earley_set_id);
Given an Earley set ID as its argument, the
earleme() recognizer method returns the corresponding earleme. Every integer in the range from 0 to the ID of the latest Earley is a valid Earley set ID, and every valid Earley set ID corresponds to an earleme. If the argument of
earleme() is greater than the latest Earley set ID,
earleme() returns Perl
There is currently no method that translates from earleme to Earley set. Earley set to earleme translation is a well-behaved one-to-one function in all input models -- for every Earley set there is a earleme, and every earleme is mapped to by at most one Earley set. Earleme to Earley set translation is far less well-behaved. In many input models, it is a partial function -- there are some earlemes that are in the valid range of earlemes but do not map to any Earley set.
Earleme to Earley set translation is often not needed. When it is, it can be implemented at the application level, with the application taking advantage of what it knows about its choice of input model.
Processes all tokens at the current earleme and advances the current earleme by 1. If the earleme cannot be completed, an exception is thrown. Otherwise, an "event" count is returned. If zero is returned, the earleme was completed without event.
In this context, an "event" is one of a list of occurrences of special interest during the successful completion of an earleme. One example of an event is parse exhaustion. To determine if one of the events was parse exhaustion, the
exhausted method can be used.
All tokens read using the
alternative method start at one location -- the current earleme. When reading input using the
earleme_complete is used to move complete processing at the current earleme and move forward in the input stream.
earleme_complete may be called even if the
alternative method has been not called since the last call to
earleme_complete. This will create an earleme with no tokens. In certain input models, such as the character-per-earleme model, this can be useful.
$recce->exhausted() and die 'Recognizer exhausted';
exhausted method returns a Perl true if parsing in a recognizer is exhausted, and a Perl false otherwise. Parsing is exhausted when the recognizer will not accept any further input. This method is seldom used. For most applications, the return values of the other methods provide sufficient information about the parse status, and there is no need to specifically test for parse exhaustion.
In the default input model, a recognizer was exhausted if the
read method returned 0, indicating that no tokens would be accepted at the new current earleme. In alternative input models,
earleme_complete will always return 0 when the parser is exhausted, but the converse is not always true:
earleme_complete may return 0 even in some cases when the parser is not exhausted.
earleme_complete returns 0, no input will be accepted at the current earleme. But in input models which allow tokens longer than 1, it is possible for input to be accepted at earlemes after the current earleme, even if no input is accepted at the current earleme.
Indicates that input is finished. Calling
end_input is not necessary or useful in the default input model, because in the default input model no token has a length greater than 1.
end_input method takes no arguments. The
end_input method returns a Perl true value on success. On failure, it throws an exception.
In alternative input models, calling the
earleme_complete method once input is finished does not ensure that all input has been processed. The
earleme_complete method completes the current earleme, but in alternative models, tokens may extend well past the current earleme. The
end_input method ensures that all input is processed.
end_input multiple times on the same recognizer object is harmless, but useless. The second and subsequent calls will return a Perl true, but will have no effect.
A recognizer can mix calls to its
read method with calls to its
alternative method. The
read method has the same effect as a single call to the
alternative method, followed immediately by a call of the
Marpa allows ambiguous tokens. Several Marpa tokens can start at a single parsing location. Ambiguous tokens can be of various lengths. Tokens can also overlap.
Potentially ambiguous lexing occurs when more than one token starts at a single earleme. When potentially ambiguous lexing occurs, it becomes possible for there to be more than one sequence of tokens.
An actual lexical ambiguity only occurs if more than one of the potential token sequences is consistent with the grammar. If there is no actual lexical ambiguity, Marpa will use the only token choice that is consistent with the grammar.
When lexing is actually ambiguous, Marpa will use all the alternatives consistent with the grammar. When the lexing in a parse is actually ambiguous, the parse will be ambiguous. From the point of view of Marpa's semantics, ambiguities caused by lexing look the same as ambiguities caused by an ambiguous grammar.
In the standard terminology, if a grammar produces more than one parse tree for any input, then that grammar must be ambiguous. In Marpa this is not strictly true. In Marpa, if the input is ambiguous, even an unambiguous grammar can produce than one parse.
A duplicate token is a token of the same type and the same length as another that was read at the same earleme. Duplicate tokens are impossible in the default, token-stream, model. This is because in the token-stream model only one token can be read at each earleme.
In alternative models, more than one token may be read at an earleme, and duplicates are possible. Marpa detects duplicate tokens and treats them as "hard errors" -- Marpa throws an exception when it sees a duplicate token. Marpa's assumption is that duplicate tokens indicate an error at the application level.
An application can retry input after a duplicate token, if it catches the exception. In the future, if recovery from duplicate tokens is found to be a useful technique, Marpa may provide an option to change its behavior, so that a soft failure is returned when there is a duplicate token.
While scanning, Marpa keeps track of the current earleme. Earlemes in a parse start at earleme 0 and increase numerically. The earleme immediately following earleme 0 is earleme 1, the earleme immediately following earleme 1 is earleme 2, and so on. The earleme immediately following earleme N is always earleme N+1.
Distance in the earleme stream is what you would intuitively expect it to be. The distance between earleme X and earleme Y is the absolute value of the difference between X and Y, |X-Y|. The distance from earleme 3 to earleme 6, for example, is 3 earlemes.
Whenever a token is given to Marpa to be scanned, it starts at the current earleme. In addition to the type and value of the token, Marpa must be told token's length in earlemes. The length of a Marpa token must be greater than zero.
This earleme length will become the distance from the start of the token to the end of the token. If the length of the token is L, and the number of the current earleme is C, the end of the token will be at the earleme whose number is C+L.
Many different models of the relationship between tokens and earlemes are possible, but two are particularly important. One is the one-token-per-earleme model, which is the default, and which has already been described. The other is the one-character-per-earleme model.
In the one-character-per-earleme model, every character will be treated as being exactly one earleme in length. If a token is more than one character in length, that token will span earlemes. When the lexing is ambiguous, tokens may overlap.
When a one-character-per-earleme model of input is used, there may be many earlemes at which no tokens start. For example, in a straightforward character-per-earleme implementation of a grammar for a language which allows comments, no tokens will start at any earlemes which corresponds to character locations inside a comment.
So far only the token-per-earleme and character-per-earleme models have seen any real use in Marpa programs. But other models are certainly possible. Using earlemes, you can structure your input in almost any way you like.
There are only three restrictions:
Scanning always starts at earleme 0.
All tokens starting at earleme N must be scanned before any tokens starting at earleme N+1. In other words, the tokens must be scanned in non-decreasing order by start earleme.
Every token must have a length, in earlemes, which is greater than zero. In other words, token length can never be zero or negative.
Copyright 2012 Jeffrey Kegler This file is part of Marpa::R2. Marpa::R2 is free software: you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. Marpa::R2 is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. You should have received a copy of the GNU Lesser General Public License along with Marpa::R2. If not, see http://www.gnu.org/licenses/.