wip -
Word Interpreter
package require Tcl 8.4
package require wip ? 2.2 ?
package require snit ? 1.3 ?
package require struct::set
::wip wipName engine arg...
def name
def name method_prefix
wipName option ? arg arg ... ?
wip::dsl ? suffix ?
wipName def name ? method_prefix ?
wipName defl names
wipName defd dict
wipName deflva name...
wipName defdva (name method_prefix)...
wipName undefl names
wipName undefva name...
wipName unknown cmdprefix
wipName runl wordlist
wipName run word...
wipName run_next
wipName run_next_while acceptable
wipName run_next_until rejected
wipName run_next_if acceptable
wipName run_next_ifnot rejected
wipName next
wipName peek
wipName peekall
wipName insertl at wordlist
wipName replacel wordlist
wipName pushl wordlist
wipName addl wordlist
wipName insert at word...
wipName replace word...
wipName push word...
wipName add word...
This package provides a micro interpreter for lists of words. Domain
specific languages based on this will have a bit of a Forth feel, with
the input stream segmented into words and any other structuring left
to whatever the language desired. Note that we have here in essence
only the core dispatch loop, and no actual commands whatsoever, making
this definitely only a Forth feel and not an actual Forth.
The idea is derived from Colin McCormack's treeql processor,
modified to require less boiler plate within the command
implementations, at the expense of, likely, execution speed. In
addition the interface between processor core and commands is more
complex too.
Word interpreters have a mappping from the names of the language
commands they shall recognize to the methods in the engine to invoke
for them, and possibly fixed arguments for these methods. This mapping
is largely static, however it is possible to change it during the
execution of a word list (= program).
At the time a language command is defined the word interpreter will
use snit's introspection capabilities to determine the
number of arguments expected by the method of the egnine, and together
with the number of fixed arguments supplied in the method prefix of
the mapping it then knows how many arguments the language command is
expecting. This is the command's arity. Variable-argument
methods (i.e. with the last argument named args) are not
allowed and will cause the word interpreter to throw an error at
definition time.
Note that while I said snit's abilities the engine object
can be written in any way, as long as it understands the method
info args, which takes a method name and returns the list
of arguments for that method.
When executing a list of words (aka program) the first word is always
taken as the name of a language command, and the next words as its
arguments, per the arity of the command. Command and argument
words are removed from the list and then associated method of the
engine is executed with the argument words. The process then repeats
using the then-first word of the list.
Note that the methods implementing the language commands may have full
access to the list of words and are allowed to manipulate as they see
fit.
-
This means, for example, that while we cannot specify
variable-argument methods directly they can consume words after their
fixed arguments before returning to the execution loop. This may be
under the control of their fixed arguments.
-
Another possibility is the use of method run_next and its
variants to execute commands coming after the current command,
changing the order of execution.
-
Execution can be further changed by use of the program accessor
methods which allow a command implementation to modify the remaining
list of words (insert, replace, prepend, append words) without
executing them immediately.
-
At last the basic run methods save and restore an existing
list of words when used, enabling recursive use from within command
implementations.
The main command of the package is:
-
::wip wipName engine arg...
-
The command creates a new word interpreter object with an associated
global Tcl command whose name is wipName. If however the string
%AUTO% was used as object name the package will generate its
own unique name for the object.
The engine is the object the word interpreter will dispatch all
recognized commands to, and the arguments are a word list which
defines an initial mapping from language words to engine methods.
The recognized language of this word list is
-
def name
-
Defines name as command of the language, to be mapped to a
method of the engine having the same name.
-
def name method_prefix
-
Defines name as command of the language, to be mapped to the
method of the engine named in the method_prefix.
The returned command may be used to invoke various operations on the
object. It has the following general form:
-
wipName option ? arg arg ... ?
-
Option and the args determine the exact behavior of the
command.
The package additionally exports the command:
-
wip::dsl ? suffix ?
-
This command is for use within snit types which wish to use one or
more wip interpreters as a component. Use within the type definition
installs most of the boilerplate needed to setup and use a word
interpreter.
It installs a component named wip, and a method
wip_setup for initializing it. This method has to be called
from within the constructor of the type using the word interpreter.
If further installs a series of procedures which make the object API
of the word interpreter directly available to the type's methods,
without having to specify the component.
Note that this does and cannot install the language to
interpret, i.e. the mapping from words to engine methods.
It is possible to instantiate multiple word interpreter components
within a type by using different suffices as arguments to the command.
In that case the name of the component changes to
'wip_$suffix', the setup command becomes
'wip_$suffix_setup' and all the procedures also get the suffix
'_$suffix'.
The following commands are possible for word interpreter objects:
-
wipName def name ? method_prefix ?
-
Defines a language command name and maps it to the method named
in the engine's method_prefix. If the method_prefix name
is not specified it is simply the name of the language command.
-
wipName defl names
-
Defines a series of language commands, specified through the list of
names, all of which are mapped to engine methods of the same
name.
-
wipName defd dict
-
Defines a series of language commands, specified through the
dictionary dict of names and method prefixes.
-
wipName deflva name...
-
As method defl, however the list of names is specified
through multiple arguments.
-
wipName defdva (name method_prefix)...
-
As method defd, however the dictionary of names and method
prefixes is specified through multiple arguments.
-
wipName undefl names
-
Removes the named series of language commands from the mapping.
-
wipName undefva name...
-
As method undefl, however the list of names is specified
through multiple arguments.
-
wipName unknown cmdprefix
-
Sets the handler for unknown words to cmdprefix. This command
prefix takes one argument, the current word, and either throws some
error, or returns the result of executing the word, as defined by the
handler. The default handler simply throws an error.
-
wipName runl wordlist
-
Treats the list of words in wordlist as a program and executes
the contained command one by one. The result of the command executed
last is returned as the result of this command.
The wordlist is stored in the object for access by the other
run-methods, and the general program accessor methods (see
below). A previously stored wordlist is saved during the execution of
this method and restored before it returns. This enables the recursive
execution of word lists within word lists.
-
wipName run word...
-
As method runl, however the list of words to execute is
specified through multiple arguments.
-
wipName run_next
-
Low-level method. Determines the next word in the list of words, and
its arguments, and then executes it. The result of the executed word
is the result of this method.
Exposed for use within command implementations.
The methods run and runl use it to execute words
until their word list is exhausted.
-
wipName run_next_while acceptable
-
Low-level method. Invokes the method run_next as long as the
next word is in the set of acceptable words, and the program is
not empty. The result of the command executed last is returned as the
result of this command.
Exposed for use within command implementations to change the order of
execution.
-
wipName run_next_until rejected
-
Low-level method. Invokes the method run_next until the next
word is in the set of rejected words, and the program is not
empty. The result of the command executed last is returned as the
result of this command.
Exposed for use within command implementations to change the order of
execution.
-
wipName run_next_if acceptable
-
Low-level method. Invokes the method run_next if the next
word is in the set of acceptable words, and the program is not
empty. The result of the command executed last is returned as the
result of this command.
Exposed for use within command implementations to change the order of
execution.
-
wipName run_next_ifnot rejected
-
Low-level method. Invokes the method run_next if the next
word is not in the set of rejected words, and the program is not
empty. The result of the command executed last is returned as the
result of this command.
Exposed for use within command implementations to change the order of
execution.
-
wipName next
-
Returns the next word in the programm. The word is also removed.
-
wipName peek
-
Returns the next word in the programm without removing it
-
wipName peekall
-
Returns the remaining programm in toto.
-
wipName insertl at wordlist
-
Basic programm accessor method. Inserts the specified wordlist
into the program, just before the word at position at. Positions
are counted from zero.
-
wipName replacel wordlist
-
Basic programm accessor method. Replaces the whole stored program with
the specified wordlist.
-
wipName pushl wordlist
-
Program accessor method. The specified wordlist is added to the
front of the remaining program. Equivalent to
$wip insertl 0 $wordlist
-
wipName addl wordlist
-
Program accessor method. The specified wordlist is appended at
the end of the remaining program. Equivalent to
$wip insertl end $wordlist
-
wipName insert at word...
-
Like method insertl, except the words are specified through
multiple arguments.
-
wipName replace word...
-
Like method setl, except the words are specified through
multiple arguments.
-
wipName push word...
-
Like method pushl, except the words are specified through
multiple arguments.
-
wipName add word...
-
Like method addl, except the words are specified through
multiple arguments.
No examples yet.
This document, and the package it describes, will undoubtedly contain
bugs and other problems.
Please report such in the category
wip 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.
word, list, interpreter