[comment {-*- tcl -*- doctools manpage}] [manpage_begin page n 1.0] [copyright {2005 Andreas Kupries }] [titledesc {Parser Generator}] [moddesc {Development Tools}] [category {Page Parser Generator}] [description] [para] The application described by this document, [syscmd page], is actually not just a parser generator, as the name implies, but a generic tool for the execution of arbitrary transformations on texts. [para] Its genericity comes through the use of [term plugins] for reading, transforming, and writing data, and the predefined set of plugins provided by Tcllib is for the generation of memoizing recursive descent parsers (aka [term {packrat parsers}]) from grammar specifications ([term {Parsing Expression Grammars}]). [para] [syscmd page] is written on top of the package [package page::pluginmgr], wrapping its functionality into a command line based application. All the other [package page::*] packages are plugin and/or supporting packages for the generation of parsers. The parsers themselves are based on the packages [package grammar::peg], [package grammar::peg::interp], and [package grammar::mengine]. [subsection {COMMAND LINE}] [list_begin definitions] [call [cmd page] [opt [arg options]...] [opt "[arg input] [opt [arg output]]"]] This is general form for calling [syscmd page]. The application will read the contents of the file [arg input], process them under the control of the specified [arg options], and then write the result to the file [arg output]. [para] If [arg input] is the string [const -] the data to process will be read from [const stdin] instead of a file. Analogously the result will be written to [const stdout] instead of a file if [arg output] is the string [const -]. A missing output or input specification causes the application to assume [const -]. [para] The detailed specifications of the recognized [arg options] are provided in section [sectref OPTIONS]. [list_begin arguments] [arg_def path input in] This argument specifies the path to the file to be processed by the application, or [const -]. The last value causes the application to read the text from [const stdin]. Otherwise it has to exist, and be readable. If the argument is missing [const -] is assumed. [arg_def path output in] This argument specifies where to write the generated text. It can be the path to a file, or [const -]. The last value causes the application to write the generated documented to [const stdout]. [para] If the file [arg output] does not exist then [lb]file dirname $output[rb] has to exist and must be a writable directory, as the application will create the fileto write to. [para] If the argument is missing [const -] is assumed. [list_end] [list_end] [subsection OPERATION] ... reading ... transforming ... writing - plugins - pipeline ... [subsection OPTIONS] This section describes all the options available to the user of the application. Options are always processed in order. I.e. of both [option --help] and [option --version] are specified the option encountered first has precedence. [para] Unknown options specified before any of the options [option -rd], [option -wr], or [option -tr] will cause processing to abort with an error. Unknown options coming in between these options, or after the last of them are assumed to always take a single argument and are associated with the last plugin option coming before them. They will be checked after all the relevant plugins, and thus the options they understand, are known. I.e. such unknown options cause error if and only if the plugin option they are associated with does not understand them, and was not superceded by a plugin option coming after. [para] Default options are used if and only if the command line did not contain any options at all. They will set the application up as a PEG-based parser generator. The exact list of options is [para] [example {-c peg}] [para] And now the recognized options and their arguments, if they have any: [para] [list_begin options] [opt_def --help] [opt_def -h] [opt_def -?] When one of these options is found on the command line all arguments coming before or after are ignored. The application will print a short description of the recognized options and exit. [opt_def --version] [opt_def -V] When one of these options is found on the command line all arguments coming before or after are ignored. The application will print its own revision and exit. [opt_def -P] This option signals the application to activate visual feedback while reading the input. [opt_def -T] This option signals the application to collect statistics while reading the input and to print them after reading has completed, before processing started. [opt_def -D] This option signals the application to activate logging in the Safe base, for the debugging of problems with plugins. [opt_def -r parser] [opt_def -rd parser] [opt_def --reader parser] These options specify the plugin the application has to use for reading the [arg input]. If the options are used multiple times the last one will be used. [opt_def -w generator] [opt_def -wr generator] [opt_def --writer generator] These options specify the plugin the application has to use for generating and writing the final [arg output]. If the options are used multiple times the last one will be used. [opt_def -t process] [opt_def -tr process] [opt_def --transform process] These options specify a plugin to run on the input. In contrast to readers and writers each use will [emph not] supersede previous uses, but add each chosen plugin to a list of transformations, either at the front, or the end, per the last seen use of either option [option -p] or [option -a]. The initial default is to append the new transformations. [opt_def -a] [opt_def --append] These options signal the application that all following transformations should be added at the end of the list of transformations. [opt_def -p] [opt_def --prepend] These options signal the application that all following transformations should be added at the beginning of the list of transformations. [opt_def --reset] This option signals the application to clear the list of transformations. This is necessary to wipe out the default transformations used. [opt_def -c file] [opt_def --configuration file] This option causes the application to load a configuration file and/or plugin. This is a plugin which in essence provides a pre-defined set of commandline options. They are processed exactly as if they have been specified in place of the option and its arguments. This means that unknown options found at the beginning of the configuration file are associated with the last plugin, even if that plugin was specified before the configuration file itself. Conversely, unknown options coming after the configuration file can be associated with a plugin specified in the file. [para] If the argument is a file which cannot be loaded as a plugin the application will assume that its contents are a list of options and their arguments, separated by space, tabs, and newlines. Options and argumentes containing spaces can be quoted via double-quotes (") and quotes ('). The quote character can be specified within in a quoted string by doubling it. Newlines in a quoted string are accepted as is. [comment {"}] [list_end] [subsection PLUGINS] [syscmd page] makes use of four different types of plugins, namely: readers, writers, transformations, and configurations. Here we provide only a basic introduction on how to use them from [syscmd page]. The exact APIs provided to and expected from the plugins can be found in the documentation for [package page::pluginmgr], for those who wish to write their own plugins. [para] Plugins are specified as arguments to the options [option -r], [option -w], [option -t], [option -c], and their equivalent longer forms. See the section [sectref OPTIONS] for reference. [para] Each such argument will be first treated as the name of a file and this file is loaded as the plugin. If however there is no file with that name, then it will be translated into the name of a package, and this package is then loaded. For each type of plugins the package management searches not only the regular paths, but a set application- and type-specific paths as well. Please see the section [sectref {PLUGIN LOCATIONS}] for a listing of all paths and their sources. [para] [list_begin definitions] [def "[option -c] [arg name]"] Configurations. The name of the package for the plugin [arg name] is "page::config::[arg name]". [para] We have one predefined plugin: [list_begin definitions] [def [emph peg]] It sets the application up as a parser generator accepting parsing expression grammars and writing a packrat parser in Tcl. The actual arguments it specifies are: [para] [example { --reset --append --reader peg --transform reach --transform use --writer me }] [para] [list_end] [def "[option -r] [arg name]"] Readers. The name of the package for the plugin [arg name] is "page::reader::[arg name]". [para] We have five predefined plugins: [list_begin definitions] [def [emph peg]] Interprets the input as a parsing expression grammar ([term PEG]) and generates a tree representation for it. Both the syntax of PEGs and the structure of the tree representation are explained in their own manpages. [def [emph hb]] Interprets the input as Tcl code as generated by the writer plugin [emph hb] and generates its tree representation. [def [emph ser]] Interprets the input as the serialization of a PEG, as generated by the writer plugin [emph ser], using the package [package grammar::peg]. [def [emph lemon]] Interprets the input as a grammar specification as understood by Richard Hipp's [term LEMON] parser generator and generates a tree representation for it. Both the input syntax and the structure of the tree representation are explained in their own manpages. [def [emph treeser]] Interprets the input as the serialization of a [package struct::tree]. It is validated as such, but nothing else. It is [emph not] assumed to be the tree representation of a grammar. [list_end] [def "[option -w] [arg name]"] Writers. The name of the package for the plugin [arg name] is "page::writer::[arg name]". [para] We have eight predefined plugins: [list_begin definitions] [def [emph identity]] Simply writes the incoming data as it is, without making any changes. This is good for inspecting the raw result of a reader or transformation. [def [emph null]] Generates nothing, and ignores the incoming data structure. [def [emph tree]] Assumes that the incoming data structure is a [package struct::tree] and generates an indented textual representation of all nodes, their parental relationships, and their attribute information. [def [emph peg]] Assumes that the incoming data structure is a tree representation of a [term PEG] or other other grammar and writes it out as a PEG. The result is nicely formatted and partially simplified (strings as sequences of characters). A pretty printer in essence, but can also be used to obtain a canonical representation of the input grammar. [def [emph tpc]] Assumes that the incoming data structure is a tree representation of a [term PEG] or other other grammar and writes out Tcl code defining a package which defines a [package grammar::peg] object containing the grammar when it is loaded into an interpreter. [def [emph hb]] This is like the writer plugin [emph tpc], but it writes only the statements which define stat expression and grammar rules. The code making the result a package is left out. [def [emph ser]] Assumes that the incoming data structure is a tree representation of a [term PEG] or other other grammar, transforms it internally into a [package grammar::peg] object and writes out its serialization. [def [emph me]] Assumes that the incoming data structure is a tree representation of a [term PEG] or other other grammar and writes out Tcl code defining a package which implements a memoizing recursive descent parser based on the match engine (ME) provided by the package [package grammar::mengine]. [list_end] [def "[option -t] [arg name]"] Transformers. The name of the package for the plugin [arg name] is "page::transform::[arg name]". [para] We have two predefined plugins: [list_begin definitions] [def [emph reach]] Assumes that the incoming data structure is a tree representation of a [term PEG] or other other grammar. It determines which nonterminal symbols and rules are reachable from start-symbol/expression. All nonterminal symbols which were not reached are removed. [def [emph use]] Assumes that the incoming data structure is a tree representation of a [term PEG] or other other grammar. It determines which nonterminal symbols and rules are able to generate a [emph finite] sequences of terminal symbols (in the sense for a Context Free Grammar). All nonterminal symbols which were not deemed useful in this sense are removed. [list_end] [list_end] [subsection {PLUGIN LOCATIONS}] The application-specific paths searched by [syscmd page] either are, or come from: [para] [list_begin enumerated] [enum] The directory [file ~/.page/plugin] [enum] The environment variable [term PAGE_PLUGINS] [enum] The registry entry [term "HKEY_LOCAL_MACHINE\\SOFTWARE\\PAGE\\PLUGINS"] [enum] The registry entry [term "HKEY_CURRENT_USER\\SOFTWARE\\PAGE\\PLUGINS"] [list_end] [para] The type-specific paths searched by [syscmd page] either are, or come from: [para] [list_begin enumerated] [enum] The directory [file ~/.page/plugin/] [enum] The environment variable [term PAGE__PLUGINS] [enum] The registry entry [term "HKEY_LOCAL_MACHINE\\SOFTWARE\\PAGE\\\\PLUGINS"] [enum] The registry entry [term "HKEY_CURRENT_USER\\SOFTWARE\\PAGE\\\\PLUGINS"] [list_end] [para] Where the placeholder [term ] is always one of the values below, properly capitalized. [list_begin enumerated] [enum] reader [enum] writer [enum] transform [enum] config [list_end] [para] The registry entries are specific to the Windows(tm) platform, all other platforms will ignore them. [para] The contents of both environment variables and registry entries are interpreted as a list of paths, with the elements separated by either colon (Unix), or semicolon (Windows). [section {BUGS, IDEAS, FEEDBACK}] This document, and the application it describes, will undoubtedly contain bugs and other problems. Please report such in the category [emph page] of the [uri {http://sourceforge.net/tracker/?group_id=12883} {Tcllib SF Trackers}]. Please also report any ideas for enhancements you may have for either application and/or documentation. [see_also page::pluginmgr] [keywords {text processing} {parser generator}] [manpage_end]