1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537
|
% $Id: intro.tex 346 2008-02-25 00:39:00Z Franz $
\chapter{Introduction\label{introduction}}
\section{Overview\label{pyclips-overview}}
This module aims to embed a fully functional CLIPS engine in Python, and
to give to the developer a more Python-compliant interface to CLIPS
without cutting down on functionalities. In fact CLIPS is compiled into
the module in its entirety, and most API functions are bound to Python
methods. However the direct bindings to the CLIPS library (implemented
as the \module{_clips} submodule) are not described here: each function
is described by an appropriate documentation string, and accessible by
means of the \function{help()} function or through the \program{pydoc}
tool. Each direct binding maps to an API provided function. For a
detailed reference\footnote{The order of parameters is changed sometimes,
in order to allow a more intuitive use of default parameters in the
Python interface: however the meaning of each parameter is described
in the function documentation string, and it should not be difficult
for the programmer to correctly understand the relationship between a
module function and the corresponding CLIPS API.} for these functions
see \clipsapg{}, available for download at the CLIPS website.
\pyclips{} is also capable of generating CLIPS text and binary files:
this allows the user to interact with sessions of the CLIPS system
itself.
An important thing to know, is that \pyclips{} implements CLIPS as a
separated\footnote{This has an impact on the way the module can be used,
as the engine is only set up once when the module is \code{import}ed the
first time.} engine: in the CLIPS module implementation, CLIPS
``lives'' in its own memory space, allocates its own objects. The module
only provides a way to send information and commands to this engine
and to retrieve results from it.
\subsection{Structure\label{pyclips-ov-structure}}
\pyclips{} is organized in a package providing several classes and
top-level functions. Also, the module provides some objects that are
already instanced and give access to some of the CLIPS internal
functions and structures, including debug status and engine I/O.
CLIPS is accessible through these classes and functions, that send
appropriate commands to the underlying engine and retrieve the
available information. Many of the CLIPS classes, constructs and
objects are shadowed by Python classes and objects. However, whereas
\pyclips{} classes provide a comfortable way to create objects that
reference the actual engine objects, there is no one-to-one
mapping between the two memory spaces: for instance, when a Python
object is deleted (via the \keyword{del} command), the corresponding
CLIPS object will still remain alive in the CLIPS memory space. An
appropriate command is necessary to remove the object from the
underlying engine, and this is provided by the module interface.
\subsection{Interactive Usage\label{pyclips-ov-interactive}}
The \pyclips{} package can also be used interactively, since it can
inspect an underlying CLIPS session and give some of the output that
CLIPS usually provides when used as an interactive shell.
A simple interactive session with \pyclips{} follows:
\begin{verbatim}
>>> import clips
>>> clips.Reset()
>>> clips.Assert("(duck)")
<Fact 'f-1': fact object at 0x00DE4AE0>
>>> clips.BuildRule("duck-rule", "(duck)", "(assert (quack))", "the Duck Rule")
<Rule 'duck-rule': defrule object at 0x00DA7E00>
>>> clips.PrintRules()
MAIN:
duck-rule
>>> clips.PrintAgenda()
MAIN:
0 duck-rule: f-1
For a total of 1 activation.
>>> clips.PrintFacts()
f-0 (initial-fact)
f-1 (duck)
For a total of 2 facts.
>>> clips.Run()
>>> clips.PrintFacts()
f-0 (initial-fact)
f-1 (duck)
f-2 (quack)
For a total of 3 facts.
\end{verbatim}
Users of the CLIPS interactive shell will find the \pyclips{} output
quite familiar. In fact the \function{Print\emph{<object>}()} functions
are provided for interactive use, and retrieve their output directly from
the underlying CLIPS engine I/O subsystem, in order to resemble an
interactive CLIPS session. Other functions are present to retrieve
object names, values and the so called \emph{pretty-print forms} for
programmatic use.
\section{Implementation Structure\label{pyclips-implstructure}}
This section describes the guidelines and considerations that lead
to this CLIPS interface implementation. For the developers which
normally use CLIPS as a development environment or expert systems
shell, the architecture of the \pyclips{} module will look a little
bit different, and in some ways could also seem confusing.
The main topics covered by these sections are the \emph{Implementation
of Constructs as Classes}, the \emph{Implementation of CLIPS I/O
Subsystem}, the \emph{Configuration and Debug Objects}, the
\emph{Coexistence of Global and Environment-Aware Engines} and the
\emph{Conventions used for Naming} which explains the rules that
helped choose the current naming scheme for classes, functions and
objects implemented in \pyclips{}.
\subsection{Implementation of Constructs as Classes\label{pyclips-ov-casc}}
CLIPS users know that this shell offers several constructs to populate
the system memory. These constructs are not described here, since a
detailed explaination of the CLIPS language can be found in the
official CLIPS documentation. These constructs, many of which have
their particular syntax, create ``objects'' (not necessarily in the
sense of OOP\footnote{In fact, the word \emph{object} is used here
to indicate an item that takes part in a program. For instance, one
such object can be a \class{Rule}: it is not a proper OO object,
but something that in imperative languages would act as a control
structure.}, although some of these can be \class{Instance}s of
\class{Class}es) in the subsystem memory.
The choice of implementing most of these constructs as classes gives
to \pyclips{} a more organic structure. Most of the construct classes
share similarities which make the interface structure simpler and the
access to CLIPS objects more systematic.
Most constructs are implemented as \emph{factory functions} which
return instances of Python classes. These Python instances (that
shadow the corresponding CLIPS objects), on their turn, have methods
and properties which operate directly on the objects they map in the
CLIPS subsystem. Methods and properties provide both access and
\emph{send messages} to these objects.
An example of this follows:
\begin{verbatim}
>>> import clips
>>> f0 = clips.Assert("(duck)")
>>> print f0
f-0
>>> print f0.Exists()
True
>>> f0.Retract()
>>> print f0.Exists()
False
\end{verbatim}
In the above example, a fact (\code{(duck)}) is asserted and then
retracted. The assertion is done by means of a module-level function
(\function{Assert()}) and the fact is retracted using a method of the
shadow object (\var{f0}). A verification on the CLIPS \code{fact} object,
using the Python \class{Fact} instance\footnote{It should be clear
that terminology differs semantically from Python system to the CLIPS
system: while OOP, to which terms used in Python are coherent, uses
the words \emph{method}, \emph{instance} and so on with a particular
meaning (to which Python developers are familiar), CLIPS terminology
often differs from OOP, sometimes only slightly but at other times more
substantially. The reader should note that throughout this manual
each term is used -- as far as possible -- with the meaning that it
assumes in its specific environment. In this case, the word
\emph{instance} represents the instance of a Python \keyword{class},
and is not referred to an entity in CLIPS.} \var{f0}'s method
\function{Exists()}, shows that after invoking the \function{Retract()}
method of \var{f0} the \code{fact} object does no longer exists in the
CLIPS subsystem, thus it has been actually retracted.
As stated previously, this does not remove the Python object
(a \class{Fact} instance) from the namespace pertinent to Python itself:
as it can be seen from the code snip shown above, \code{f0} is still a
functional instance, and can be queried about the existence of the
corresponding object in CLIPS.
Objects in the CLIPS operating space can be referenced by more than a
Python object (or even not be referenced at all, if CLIPS creation does
not correspond to a Python assignment), as demonstrated by the following
code:
\begin{verbatim}
>>> clips.Reset()
>>> f1 = clips.Assert("(duck)")
>>> clips.Assert("(quack)")
<Fact 'f-2': fact object at 0x00DE8420>
>>> f1
<Fact 'f-1': fact object at 0x00DE3020>
>>> fl = clips.FactList()
>>> f1b = fl[1]
>>> f1b
<Fact 'f-1': fact object at 0x00E08C40>
\end{verbatim}
Both \var{f1} and \var{f1b} refer to the same object in CLIPS namespace,
but their address is different: in fact they are two different Python
objects (equality test fails) but correspond to the same \code{fact} in
CLIPS.
\begin{seealso}
\sclipsbpg{}
\end{seealso}
\subsection{Implementation of CLIPS I/O Subsystem\label{pyclips-ov-io}}
The CLIPS shell interacts with the user answering to typed in commands
with some informational output. An interactive CLIPS session will show
this:
\begin{verbatim}
CLIPS> (reset)
CLIPS> (watch activations)
CLIPS> (defrule duck-rule "the Duck Rule"
(duck)
=>
(assert (quack)))
CLIPS> (assert (duck))
==> Activation 0 duck-rule: f-1
<Fact-1>
CLIPS> (run)
CLIPS> (facts)
f-0 (initial-fact)
f-1 (duck)
f-2 (quack)
For a total of 3 facts.
\end{verbatim}
Each time a \code{fact} is asserted, CLIPS outputs a string containing
its index, and since we decided to show some debug output about
activations, CLIPS produces a line as soon as \code{duck} is asserted,
since \code{duck-rule} would be activated by this. Although in an
interactive session all of the output would go to the terminal, CLIPS
logically considers the ``streams'' for different output types as
separated: in fact, debug output (the one generated by the \code{watch}
command) goes to a special stream called \code{wtrace}. In this special
case, for instance, the debug output can be captured by \pyclips{}
through a special stream-like Python object, which provides a
\function{Read()} function\footnote{Note that \function{Read()} is
capitalized: this is because the stream-like objects do not really act
as ``files'' as many objects which can be read in Python do. So it
becomes impossible to use these \pyclips{} objects where a file-like
object is to be used.}. Comparing the behaviour of two interactive
sessions, the former in the CLIPS subsystem and the latter in Python,
will help to understand the close relationship between CLIPS I/O and
\pyclips{} stream objects. CLIPS will interact with the user as
follows:
\begin{verbatim}
CLIPS> (defrule sayhello
(hello)
=>
(printout t "hello, world!" crlf))
CLIPS> (assert (hello))
==> Activation 0 sayhello: f-1
<Fact-1>
CLIPS> (run)
hello, world!
\end{verbatim}
And the Python counterpart follows:
\begin{verbatim}
>>> import clips
>>> clips.DebugConfig.ActivationsWatched = True
>>> r0 = clips.BuildRule("sayhello", "(hello)",
'(printout stdout "hello, world!" crlf)')
>>> print r0.PPForm()
(defrule MAIN::sayhello
(hello)
=>
(printout stdout "hello, world!" crlf))
>>> clips.Assert("(hello)")
<Fact 'f-0': fact object at 0x00DE81C0>
>>> t = clips.TraceStream.Read()
>>> print t
==> Activation 0 sayhello: f-0
>>> clips.Run()
>>> t = clips.StdoutStream.Read()
>>> print t
hello, world!
\end{verbatim}
The I/O access objects can be used both in interactive and unattended
sessions. In this case, they can be useful to retrieve periodical
information about CLIPS internal status, since most of the output
provided can be easily interpreted in a programmatic way. Also, there
is one more stream called \var{StdinStream} (which has a
\function{Write()} method) that might be useful to send input to the
CLIPS engine when some ``user interaction'' is required\footnote{This only
happens actually when CLIPS invokes the \code{read} or \code{readline}
functions.}.
There is no way to create other instances of these streams: the
high-level module hides the class used to build these objects. This
is because the I/O streams have to be considered like ``physical
devices'' whose use is reserved to the engine to report trace and
debug information as well as user requested output.
These I/O streams will be described later in the detail, since each
one can be used to report about a specific task.
\note{The streams have to be explicitly read: there is no way to
receive a notification from CLIPS that some output has been written.
In other words, the \pyclips{} engine is not \emph{event driven} in
its interaction with Python.}
\subsection{Configuration and Debug Objects\label{pyclips-ov-cado}}
As well as I/O streams, there are two other objects directly provided
by \pyclips{}. These objects provide access to the CLIPS engine global
configuration. Many aspects of the CLIPS engine, that in the command
line environment would be configured using particular commands, are
accessible via the \var{EngineConfig} (global engine configuration)
object and the \var{DebugConfig} (global debug and trace configuration)
object. For example, we can take the code snip shown above:
\begin{verbatim}
>>> clips.DebugConfig.ActivationsWatched = True
\end{verbatim}
(...)
\begin{verbatim}
>>> t = clips.TraceStream.Read()
>>> print t
==> Activation 0 sayhello: f-0
\end{verbatim}
The \code{clips.DebugConfig.ActivationsWatched = True} line tells to the
underlying subsystem that debug information about \emph{rule activations}
has to be written to the proper stream (the stream dedicated to debug
output in CLIPS is called \code{wtrace} and is accessible in \pyclips{}
through the \var{TraceStream} object).
As it has been said for the I/O streams, these objects cannot be instanced
by the user: access to these objects affects global (or at least
\emph{environmental}, we will see the difference later) configuration,
so it would be of no meaning for the user to create more, possibly
confusing, instances of such objects.
\subsection{Coexistence of Global and Environment-Aware Engines\label{pyclips-ov-env}}
As of version 6.20, CLIPS API offers the possibility to have several
\emph{environments} in which to operate. We can consider environments as
separate engines that only share the operating mode, in other words
``the code''. \pyclips{} also implements environments by means of a special
\class{Environment} class. This class implements all the features
provided by the top level methods and classes. The \class{Environment}
class reimplements all classes provided by \pyclips{}, but -- although
their behaviour is quite similar -- methods of classes provided by
\class{Environment} only affect the CLIPS environment represented by the
\class{Environment} instance itself.
There is normally no need to use environments. However, access to them
is provided for CLIPS ``gurus'' who want to have more than one engine
working at the same time. The end user of \pyclips{} will see no real
difference between a call to a function and its environmental counterpart
(defined as \emph{companion function} in the official CLIPS documentation),
apart from being called as a member function of an \class{Environment}
object.
A simple example will be explanatory:
\begin{verbatim}
>>> clips.Clear()
>>> clips.Reset()
>>> e0 = clips.Environment()
>>> e1 = clips.Environment()
>>> e0.Assert("(duck)")
<Fact 'f-0': fact object at 0x00E7D960>
>>> e1.Assert("(quack)")
<Fact 'f-0': fact object at 0x00E82220>
>>> e0.PrintFacts()
f-0 (duck)
For a total of 1 fact.
>>> e1.PrintFacts()
f-0 (quack)
For a total of 1 fact.
\end{verbatim}
\subsection{Using External Functions in CLIPS\label{pyclips-ov-extfuncs}}
\pyclips{} gives the ability to users to call Python code from within
the CLIPS subsystem. Virtually every function defined in Python can be
called from CLIPS code using the special CLIPS function \code{python-call}.
However, since CLIPS has different basic types than Python, in most cases
it would be useful for modules that implement function to be called in
the CLIPS engine to import the \pyclips{} module themselves, in order to
be aware of the structures that CLIPS uses.
Functions have to be registered in \pyclips{} in order to be available
to the underlying engine, and the registration process can dynamically
occur at any moment.
A simple example follows:
\begin{verbatim}
>>> import clips
>>> def py_square(x):
return x * x
>>> clips.RegisterPythonFunction(py_square)
>>> print clips.Eval("(python-call py_square 7)")
49
>>> print clips.Eval("(python-call py_square 0.7)")
0.49
\end{verbatim}
A more detailed description of the features provided by \code{python-call}
can be found in the appendices.
\subsection{Conventions Used for Naming\label{pyclips-ov-names}}
In \pyclips{}, the simple convention that is used is that all valuable
content exposed has a name beginning with a capital letter. Names
beginning with a single underscore have normally no meaning for the
\pyclips{} user. Functions, class names and objects use mixed capitals
(as in Java), and \emph{manifest constants} (names used in \emph{lieu}
of explicit values to pass instructions to CLIPS functions or
properties) are all capitalized, as is usual for the C language.
CLIPS users will perhaps be confused because often the constructs in
CLIPS are expressed by keywords containing a \code{def} prefix. The
choice was made in \pyclips{} to drop this prefix in many cases: the use
of this prefix has a strong logic in the CLIPS language, because in this
way the developer knows that a \emph{construct} is used, that is, a
\emph{definition} is made. The keyword used to instance this definition,
both encapsulates the meaning of \code{def}inition itself, and also
the type of construct that is being defined (e.g. \code{def}ine a
\code{rule} is \code{defrule}), thus avoiding making constructs more
difficult by means of two separate keywords. In \pyclips{}, since the
definition happens at class declaration and the instantiation of classes
shadows a construct definition when it has already been performed, it
seemed unnecessary to keep the prefix: in fact, to follow the above
example, it does not seem correct to refer to a rule within the CLIPS
subsystem as a ``\class{Defrule}'' object, hence it is simply referred
to as a \class{Rule}.
\subsection{Pickling Errors\label{pyclips-ov-pickle}}
Python objects cannot be pickled or unpickled. This is because, since
pickling an object would save a reference to a CLIPS entity -- which is
useless across different \pyclips{} sessions -- the unpickling process
would feed the underlying engine in an unpredictable way, or at least
would reference memory locations corresponding to previous CLIPS entities
without the engine having them allocated.
One better way to achieve a similar goal is to use the \function{Save()}
or \function{BSave()} (and related \function{Load()} or \function{BLoad()})
to save the engine\footnote{The mentioned functions are also \emph{members}
of the \class{Environment} class, in which case the \class{Environment}
status is saved.} status in its entirety.
If a single entity is needed, its \emph{pretty-print form} can be used
in most cases to recreate it using the \function{Build()} functions.
\section{Other Usage Modes\label{pyclips-ov-otheruses}}
It is also interesting that, by using some particular functions and the
provided I/O subsystem, even ``pure'' CLIPS programs can be executed by
\pyclips{}, and while the simple output from CLIPS can be read to obtain
feedback, the possibility of inspecting the internal CLIPS subsystem
state remains.
The following example, taken from the CLIPS website\footnote{In fact
the file has been slightly reformatted for typesetting reasons.},
illustrates this: first we take a full CLIPS program, saved as
\file{zebra.clp}, and reported below:
\verbatiminput{zebra.clp}
then we execute all commands (using the \function{BatchStar()} function)
in the current \class{Environment} of an interactive \pyclips{} session:
\begin{verbatim}
>>> clips.BatchStar("zebra.clp")
>>> clips.Reset()
>>> clips.Run()
>>> s = clips.StdoutStream.Read()
>>> print s
There are five houses, each of a different color, inhabited by men of
different nationalities, with different pets, drinks, and cigarettes.
The Englishman lives in the red house. The Spaniard owns the dog.
The ivory house is immediately to the left of the green house, where
the coffee drinker lives. The milk drinker lives in the middle house.
The man who smokes Old Golds also keeps snails. The Ukrainian drinks
tea. The Norwegian resides in the first house on the left. The
Chesterfields smoker lives next door to the fox owner. The Lucky
Strike smoker drinks orange juice. The Japanese smokes Parliaments.
The horse owner lives next to the Kools smoker, whose house is yellow.
The Norwegian lives next to the blue house.
Now, who drinks water? And who owns the zebra?
HOUSE | Nationality | Color | Pet | Drink | Smokes
--------------------------------------------------------------------
1 | norwegian | yellow | fox | water | kools
2 | ukrainian | blue | horse | tea | chesterfields
3 | englishman | red | snails | milk | old-golds
4 | spaniard | ivory | dog | orange-juice | lucky-strikes
5 | japanese | green | zebra | coffee | parliaments
>>> clips.PrintFacts()
f-0 (initial-fact)
f-26 (avh (a smokes) (v parliaments) (h 1))
f-27 (avh (a smokes) (v parliaments) (h 2))
f-28 (avh (a smokes) (v parliaments) (h 3))
\end{verbatim}
[... a long list of facts ...]
\begin{verbatim}
f-150 (avh (a color) (v red) (h 5))
For a total of 126 facts.
>>> li = clips.FactList()
>>> for x in li:
... if str(x) == 'f-52':
... f52 = x
>>> f52
<Fact 'f-52': fact object at 0x00E6AA10>
>>> print f52.PPForm()
f-52 (avh (a drink) (v tea) (h 2))
\end{verbatim}
You can just copy the program above to a file, say \file{zebra.clp} as
in the example, and follow the same steps to experiment with \pyclips{}
objects and with the CLIPS subsystem.
|