pt(3tcl) Parser Tools pt(3tcl)
______________________________________________________________________________
NAME
pt - Parser Tools Application
SYNOPSIS
package require Tcl 8.5
pt generate resultformat ?options...? resultfile inputformat inputfile
______________________________________________________________________________
DESCRIPTION
Are you lost ? Do you have trouble understanding this document ? In
that case please read the overview provided by the Introduction to
Parser Tools. This document is the entrypoint to the whole system the
current package is a part of.
This document describes pt, the main application of the module, a
parser generator. Its intended audience are people who wish to create a
parser for some language of theirs. Should you wish to modify the ap-
plication instead, please see the section about the application's In-
ternals for the basic references.
It resides in the User Application Layer of Parser Tools.
IMAGE: arch_user_app
COMMAND LINE
pt generate resultformat ?options...? resultfile inputformat inputfile
This sub-command of the application reads the parsing expression
grammar stored in the inputfile in the format inputformat, con-
verts it to the resultformat under the direction of the (format-
specific) set of options specified by the user and stores the
result in the resultfile.
The inputfile has to exist, while the resultfile may be created,
overwriting any pre-existing content of the file. Any missing
directory in the path to the resultfile will be created as well.
The exact form of the result for, and the set of options sup-
ported by the known result-formats, are explained in the upcom-
ing sections of this document, with the list below providing an
index mapping between format name and its associated section. In
alphabetical order:
c A resultformat. See section C Parser.
container
A resultformat. See section Grammar Container.
critcl A resultformat. See section C Parser Embedded In Tcl.
json A input- and resultformat. See section JSON Grammar Ex-
change.
oo A resultformat. See section TclOO Parser.
peg A input- and resultformat. See section PEG Specification
Language.
snit A resultformat. See section Snit Parser.
Of the seven possible results four are parsers outright (c, critcl, oo,
and snit), one (container) provides code which can be used in conjunc-
tion with a generic parser (also known as a grammar interpreter), and
the last two (json and peg) are doing double-duty as input formats, al-
lowing the transformation of grammars for exchange, reformatting, and
the like.
The created parsers fall into three categories:
.nf + --- C ---> critcl, c | + --- specialized -+ | |
---+ + --- Tcl -> snit, oo | + --- interpreted (Tcl)
------> container .fi
Specialized parsers implemented in C
The fastest parsers are created when using the result formats c
and critcl. The first returns the raw C code for the parser,
while the latter wraps it into a Tcl package using CriTcl.
This makes the latter much easier to use than the former. On the
other hand, the former can be adapted to the users' requirements
through a multitude of options, allowing for things like usage
of the parser outside of a Tcl environment, something the critcl
format doesn't support. As such the c format is meant for more
advanced users, or users with special needs.
A disadvantage of all the parsers in this section is the need to
run them through a C compiler to make them actually executable.
This is not something everyone has the necessary tools for. The
parsers in the next section are for people under such restric-
tions.
Specialized parsers implemented in Tcl
As the parsers in this section are implemented in Tcl they are
quite a bit slower than anything from the previous section. On
the other hand this allows them to be used in pure-Tcl environ-
ments, or in environments which allow only a limited set of bi-
nary packages. In the latter case it will be advantageous to
lobby for the inclusion of the C-based runtime support (notes
below) into the environment to reduce the impact of Tcl's on the
speed of these parsers.
The relevant formats are snit and oo. Both place their result
into a Tcl package containing a snit::type, or TclOO class re-
spectively.
Of the supporting runtime, which is the package pt::rde, the
user has to know nothing but that it does exist and that the
parsers are dependent on it. Knowledge of the API exported by
the runtime for the parsers' consumption is not required by the
parsers' users.
Interpreted parsing implemented in Tcl
The last category, grammar interpretation. This means that an
interpreter for parsing expression grammars takes the descrip-
tion of the grammar to parse input for, and uses it guide the
parsing process. This is the slowest of the available options,
as the interpreter has to continually run through the configured
grammar, whereas the specialized parsers of the previous sec-
tions have the relevant knowledge about the grammar baked into
them.
The only places where using interpretation make sense is where
the grammar for some input may be changed interactively by the
user, as the interpretation allows for quick turnaround after
each change, whereas the previous methods require the generation
of a whole new parser, which is not as fast. On the other hand,
wherever the grammar to use is fixed, the previous methods are
much more advantageous as the time to generate the parser is mi-
nuscule compared to the time the parser code is in use.
The relevant result format is container. It (quickly) generates
grammar descriptions (instead of a full parser) which match the
API expected by ParserTools' grammar interpreter. The latter is
provided by the package pt::peg::interp.
All the parsers generated by critcl, snit, and oo, and the grammar in-
terpreter share a common API for access to the actual parsing function-
ality, making them all plug-compatible. It is described in the Parser
API specification document.
PEG SPECIFICATION LANGUAGE
peg, a language for the specification of parsing expression grammars is
meant to be human readable, and writable as well, yet strict enough to
allow its processing by machine. Like any computer language. It was de-
fined to make writing the specification of a grammar easy, something
the other formats found in the Parser Tools do not lend themselves too.
For either an introduction to or the formal specification of the lan-
guage, please go and read the PEG Language Tutorial.
When used as a result-format this format supports the following op-
tions:
-file string
The value of this option is the name of the file or other entity
from which the grammar came, for which the command is run. The
default value is unknown.
-name string
The value of this option is the name of the grammar we are pro-
cessing. The default value is a_pe_grammar.
-user string
The value of this option is the name of the user for which the
command is run. The default value is unknown.
-template string
The value of this option is a string into which to put the gen-
erated text and the values of the other options. The various lo-
cations for user-data are expected to be specified with the
placeholders listed below. The default value is "@code@".
@user@ To be replaced with the value of the option -user.
@format@
To be replaced with the the constant PEG.
@file@ To be replaced with the value of the option -file.
@name@ To be replaced with the value of the option -name.
@code@ To be replaced with the generated text.
JSON GRAMMAR EXCHANGE
The json format for parsing expression grammars was written as a data
exchange format not bound to Tcl. It was defined to allow the exchange
of grammars with PackRat/PEG based parser generators for other lan-
guages.
For the formal specification of the JSON grammar exchange format,
please go and read The JSON Grammar Exchange Format.
When used as a result-format this format supports the following op-
tions:
-file string
The value of this option is the name of the file or other entity
from which the grammar came, for which the command is run. The
default value is unknown.
-name string
The value of this option is the name of the grammar we are pro-
cessing. The default value is a_pe_grammar.
-user string
The value of this option is the name of the user for which the
command is run. The default value is unknown.
-indented boolean
If this option is set the system will break the generated JSON
across lines and indent it according to its inner structure,
with each key of a dictionary on a separate line.
If the option is not set (the default), the whole JSON object
will be written on a single line, with minimum spacing between
all elements.
-aligned boolean
If this option is set the system will ensure that the values for
the keys in a dictionary are vertically aligned with each other,
for a nice table effect. To make this work this also implies
that -indented is set.
If the option is not set (the default), the output is formatted
as per the value of indented, without trying to align the values
for dictionary keys.
C PARSER EMBEDDED IN TCL
The critcl format is executable code, a parser for the grammar. It is a
Tcl package with the actual parser implementation written in C and em-
bedded in Tcl via the critcl package.
This result-format supports the following options:
-file string
The value of this option is the name of the file or other entity
from which the grammar came, for which the command is run. The
default value is unknown.
-name string
The value of this option is the name of the grammar we are pro-
cessing. The default value is a_pe_grammar.
-user string
The value of this option is the name of the user for which the
command is run. The default value is unknown.
-class string
The value of this option is the name of the class to generate,
without leading colons. The default value is CLASS.
For a simple value X without colons, like CLASS, the parser com-
mand will be X::X. Whereas for a namespaced value X::Y the
parser command will be X::Y.
-package string
The value of this option is the name of the package to generate.
The default value is PACKAGE.
-version string
The value of this option is the version of the package to gener-
ate. The default value is 1.
C PARSER
The c format is executable code, a parser for the grammar. The parser
implementation is written in C and can be tweaked to the users' needs
through a multitude of options.
The critcl format, for example, is implemented as a canned configura-
tion of these options on top of the generator for c.
This result-format supports the following options:
-file string
The value of this option is the name of the file or other entity
from which the grammar came, for which the command is run. The
default value is unknown.
-name string
The value of this option is the name of the grammar we are pro-
cessing. The default value is a_pe_grammar.
-user string
The value of this option is the name of the user for which the
command is run. The default value is unknown.
-template string
The value of this option is a string into which to put the gen-
erated text and the other configuration settings. The various
locations for user-data are expected to be specified with the
placeholders listed below. The default value is "@code@".
@user@ To be replaced with the value of the option -user.
@format@
To be replaced with the the constant C/PARAM.
@file@ To be replaced with the value of the option -file.
@name@ To be replaced with the value of the option -name.
@code@ To be replaced with the generated Tcl code.
The following options are special, in that they will occur
within the generated code, and are replaced there as well.
@statedecl@
To be replaced with the value of the option state-decl.
@stateref@
To be replaced with the value of the option state-ref.
@strings@
To be replaced with the value of the option string-var-
name.
@self@ To be replaced with the value of the option self-command.
@def@ To be replaced with the value of the option fun-quali-
fier.
@ns@ To be replaced with the value of the option namespace.
@main@ To be replaced with the value of the option main.
@prelude@
To be replaced with the value of the option prelude.
-state-decl string
A C string representing the argument declaration to use in the
generated parsing functions to refer to the parsing state. In
essence type and argument name. The default value is the string
RDE_PARAM p.
-state-ref string
A C string representing the argument named used in the generated
parsing functions to refer to the parsing state. The default
value is the string p.
-self-command string
A C string representing the reference needed to call the gener-
ated parser function (methods ...) from another parser fonction,
per the chosen framework (template). The default value is the
empty string.
-fun-qualifier string
A C string containing the attributes to give to the generated
functions (methods ...), per the chosen framework (template).
The default value is static.
-namespace string
The name of the C namespace the parser functions (methods, ...)
shall reside in, or a general prefix to add to the function
names. The default value is the empty string.
-main string
The name of the main function (method, ...) to be called by the
chosen framework (template) to start parsing input. The default
value is __main.
-string-varname string
The name of the variable used for the table of strings used by
the generated parser, i.e. error messages, symbol names, etc.
The default value is p_string.
-prelude string
A snippet of code to be inserted at the head of each generated
parsing function. The default value is the empty string.
-indent integer
The number of characters to indent each line of the generated
code by. The default value is 0.
-comments boolean
A flag controlling the generation of code comments containing
the original parsing expression a parsing function is for. The
default value is on.
SNIT PARSER
The snit format is executable code, a parser for the grammar. It is a
Tcl package holding a snit::type, i.e. a class, whose instances are
parsers for the input grammar.
This result-format supports the following options:
-file string
The value of this option is the name of the file or other entity
from which the grammar came, for which the command is run. The
default value is unknown.
-name string
The value of this option is the name of the grammar we are pro-
cessing. The default value is a_pe_grammar.
-user string
The value of this option is the name of the user for which the
command is run. The default value is unknown.
-class string
The value of this option is the name of the class to generate,
without leading colons. Note, it serves double-duty as the name
of the package to generate too, if option -package is not speci-
fied, see below. The default value is CLASS, applying if nei-
ther option -class nor -package were specified.
-package string
The value of this option is the name of the package to generate,
without leading colons. Note, it serves double-duty as the name
of the class to generate too, if option -class is not specified,
see above. The default value is PACKAGE, applying if neither
option -package nor -class were specified.
-version string
The value of this option is the version of the package to gener-
ate. The default value is 1.
TCLOO PARSER
The oo format is executable code, a parser for the grammar. It is a Tcl
package holding a TclOO class, whose instances are parsers for the in-
put grammar.
This result-format supports the following options:
-file string
The value of this option is the name of the file or other entity
from which the grammar came, for which the command is run. The
default value is unknown.
-name string
The value of this option is the name of the grammar we are pro-
cessing. The default value is a_pe_grammar.
-user string
The value of this option is the name of the user for which the
command is run. The default value is unknown.
-class string
The value of this option is the name of the class to generate,
without leading colons. Note, it serves double-duty as the name
of the package to generate too, if option -package is not speci-
fied, see below. The default value is CLASS, applying if nei-
ther option -class nor -package were specified.
-package string
The value of this option is the name of the package to generate,
without leading colons. Note, it serves double-duty as the name
of the class to generate too, if option -class is not specified,
see above. The default value is PACKAGE, applying if neither
option -package nor -class were specified.
-version string
The value of this option is the version of the package to gener-
ate. The default value is 1.
GRAMMAR CONTAINER
The container format is another form of describing parsing expression
grammars. While data in this format is executable it does not consti-
tute a parser for the grammar. It always has to be used in conjunction
with the package pt::peg::interp, a grammar interpreter.
The format represents grammars by a snit::type, i.e. class, whose in-
stances are API-compatible to the instances of the pt::peg::container
package, and which are preloaded with the grammar in question.
This result-format supports the following options:
-file string
The value of this option is the name of the file or other entity
from which the grammar came, for which the command is run. The
default value is unknown.
-name string
The value of this option is the name of the grammar we are pro-
cessing. The default value is a_pe_grammar.
-user string
The value of this option is the name of the user for which the
command is run. The default value is unknown.
-mode bulk|incremental
The value of this option controls which methods of pt::peg::con-
tainer instances are used to specify the grammar, i.e. preload
it into the container. There are two legal values, as listed be-
low. The default is bulk.
bulk In this mode the methods start, add, modes, and rules are
used to specify the grammar in a bulk manner, i.e. as a
set of nonterminal symbols, and two dictionaries mapping
from the symbols to their semantic modes and parsing ex-
pressions.
This mode is the default.
incremental
In this mode the methods start, add, mode, and rule are
used to specify the grammar piecemal, with each nontermi-
nal having its own block of defining commands.
-template string
The value of this option is a string into which to put the gen-
erated code and the other configuration settings. The various
locations for user-data are expected to be specified with the
placeholders listed below. The default value is "@code@".
@user@ To be replaced with the value of the option -user.
@format@
To be replaced with the the constant CONTAINER.
@file@ To be replaced with the value of the option -file.
@name@ To be replaced with the value of the option -name.
@mode@ To be replaced with the value of the option -mode.
@code@ To be replaced with the generated code.
EXAMPLE
In this section we are working a complete example, starting with a PEG
grammar and ending with running the parser generated from it over some
input, following the outline shown in the figure below:
IMAGE: flow
Our grammar, assumed to the stored in the file "calculator.peg" is
PEG calculator (Expression)
Digit <- '0'/'1'/'2'/'3'/'4'/'5'/'6'/'7'/'8'/'9' ;
Sign <- '-' / '+' ;
Number <- Sign? Digit+ ;
Expression <- Term (AddOp Term)* ;
MulOp <- '*' / '/' ;
Term <- Factor (MulOp Factor)* ;
AddOp <- '+'/'-' ;
Factor <- '(' Expression ')' / Number ;
END;
From this we create a snit-based parser via
pt generate snit calculator.tcl -class calculator -name calculator peg calculator.peg
which leaves us with the parser package and class written to the file
"calculator.tcl". Assuming that this package is then properly in-
stalled in a place where Tcl can find it we can now use this class via
a script like
package require calculator
lassign $argv input
set channel [open $input r]
set parser [calculator]
set ast [$parser parse $channel]
$parser destroy
close $channel
... now process the returned abstract syntax tree ...
where the abstract syntax tree stored in the variable will look like
set ast {Expression 0 4
{Factor 0 4
{Term 0 2
{Number 0 2
{Digit 0 0}
{Digit 1 1}
{Digit 2 2}
}
}
{AddOp 3 3}
{Term 4 4
{Number 4 4
{Digit 4 4}
}
}
}
}
assuming that the input file and channel contained the text
120+5
A more graphical representation of the tree would be
.nf +- Digit 0 0 | 1 | | +- Term 0 2 --- Number 0 2 -+-
Digit 1 1 | 2 | | | |
+- Digit 2 2 | 0 | | Expression
0 4 --- Factor 0 4 -+----------------------------- AddOp 3 3 | + |
| +- Term 4 4 --- Number 4 4 --- Digit 4 4 | 5 .fi
Regardless, at this point it is the user's responsibility to work with
the tree to reach whatever goal she desires. I.e. analyze it, transform
it, etc. The package pt::ast should be of help here, providing commands
to walk such ASTs structures in various ways.
One important thing to note is that the parsers used here return a data
structure representing the structure of the input per the grammar un-
derlying the parser. There are no callbacks during the parsing process,
i.e. no parsing actions, as most other parsers will have.
Going back to the last snippet of code, the execution of the parser for
some input, note how the parser instance follows the specified Parser
API.
INTERNALS
This section is intended for users of the application which wish to
modify or extend it. Users only interested in the generation of parsers
can ignore it.
The main functionality of the application is encapsulated in the pack-
age pt::pgen. Please read it for more information.
BUGS, IDEAS, FEEDBACK
This document, and the package it describes, will undoubtedly contain
bugs and other problems. Please report such in the category pt of the
Tcllib Trackers [http://core.tcl.tk/tcllib/reportlist]. Please also
report any ideas for enhancements you may have for either package
and/or documentation.
When proposing code changes, please provide unified diffs, i.e the out-
put of diff -u.
Note further that attachments are strongly preferred over inlined
patches. Attachments can be made by going to the Edit form of the
ticket immediately after its creation, and then using the left-most
button in the secondary navigation bar.
KEYWORDS
EBNF, LL(k), PEG, TDPL, context-free languages, expression, grammar,
matching, parser, parsing expression, parsing expression grammar, push
down automaton, recursive descent, state, top-down parsing languages,
transducer
CATEGORY
Parsing and Grammars
COPYRIGHT
Copyright (c) 2009 Andreas Kupries <andreas_kupries@users.sourceforge.net>
tcllib 1 pt(3tcl)