grammar::fa::dexec -
Execute deterministic finite automatons
package require Tcl 8.4
package require snit
package require grammar::fa::dexec ? 0.2 ?
::grammar::fa::dexec daName fa ? -any any ? ? -command cmdprefix ?
daName option ? arg arg ... ?
daName destroy
daName put symbol
daName reset
daName state
cmdprefix error code message
cmdprefix final stateid
cmdprefix reset
cmdprefix state stateid
This package provides a class for executors constructed from
deterministic finite automatons (DFA). Executors are objects
which are given a string of symbols in a piecemal fashion, perform
state transitions and report back when they enter a final state, or
find an error in the input.
For the actual creation of the DFAs the executors are based on we have
the packages grammar::fa and grammar::fa::op.
The objects follow a push model. Symbols are pushed into the executor,
and when something important happens, i.e. error occurs, a state transition,
or a final state is entered this will be reported via the callback
specified via the option -command. Note that conversion of
this into a pull model where the environment retrieves messages from
the object and the object uses a callback to ask for more symbols is
a trivial thing.
Side note:
The acceptor objects provided by grammar::fa::dacceptor
could have been implemented on top of the executors provided here, but
were not, to get a bit more performance (we avoid a number of method
calls and the time required for their dispatch).
The package exports the API described here.
-
::grammar::fa::dexec daName fa ? -any any ? ? -command cmdprefix ?
-
Creates a new deterministic executor with an associated global Tcl
command whose name is daName. This command may be used to invoke
various operations on the executor. It has the following general form:
-
daName option ? arg arg ... ?
-
Option and the args determine the exact behavior of the
command. See section EXECUTOR METHODS for more
explanations.
The executor will be based on the deterministic finite automaton
stored in the object fa. It will keep a copy of the relevant
data of the FA in its own storage, in a form easy to use for its
purposes. This also means that changes made to the fa after the
construction of the executor will not influence the executor.
If any has been specified, then the executor will convert all
symbols in the input which are unknown to the base FA to that symbol
before proceeding with the processing.
All executors provide the following methods for their manipulation:
-
daName destroy
-
Destroys the automaton, including its storage space and associated
command.
-
daName put symbol
-
Takes the current state of the executor and the symbol and
performs the appropriate state transition. Reports any errors
encountered via the command callback, as well as entering a final
state of the underlying FA.
When an error is reported all further invokations of put will
do nothing, until the error condition has been cleared via an
invokation of method reset.
-
daName reset
-
Unconditionally sets the executor into the start state of the
underlying FA. This also clears any error condition put may
have encountered.
-
daName state
-
Returns the current state of the underlying FA. This allow for
introspection without the need to pass data from the callback command.
The callback command
cmdprefix given to an executor via the
option
-command will be executed by the object at the global
level, using the syntax described below. Note that
cmdprefix is
not simply the name of a command, but a full command prefix. In other
words it may contain additional fixed argument words beyond the
command word.
-
cmdprefix error code message
-
The executor has encountered an error, and message contains a
human-readable text explaining the nature of the problem.
The code on the other hand is a fixed machine-readable text.
The following error codes can be generated by executor objects.
- BADSYM
-
An unknown symbol was found in the input. This can happen if and only
if no -any symbol was specified.
- BADTRANS
-
The underlying FA has no transition for the current combination of
input symbol and state. In other words, the executor was not able to
compute a new state for this combination.
-
cmdprefix final stateid
-
The executor has entered the final state stateid.
-
cmdprefix reset
-
The executor was reset.
-
cmdprefix state stateid
-
The FA changed state due to a transition. stateid is the new state.
This document, and the package it describes, will undoubtedly contain
bugs and other problems.
Please report such in the category
grammar_fa of the
http://sourceforge.net/tracker/?group_id=12883.
Please also report any ideas for enhancements you may have for either
package and/or documentation.
execution, running, grammar, automaton, finite automaton, state, regular expression, regular grammar, regular languages, parsing, transducer