Windows Installer Issue

Justin Kulm — Fri, 09 Nov 2018 19:51:17 -0000

There may be a trivial answer to this, however, I can't seem to install PyCLIPS via the Windows Executable available for download. I receive the following error despite having both a 32-bit and 64-bit version of Python 2.7x. I have also tried installing via pip with no luck.

Error in factmngr.c in CLIPSSrc.zip

wtdoor — Mon, 15 Jan 2018 20:32:25 -0000

My apologies; meant to put this in the Help forum. Additionally, this error seems to be address in version 6.30 of the CLIPS software distribution. Thanks!

Error in factmngr.c in CLIPSSrc.zip

wtdoor — Mon, 15 Jan 2018 20:18:19 -0000

Hello --

I believe I have finally discovered the source of the segmentation fault I alluded to in the topic "EnvIncrementFactCount call in clips_fact_assign". There appears to be an error in the factmngr.c distributed in CLIPSSrc.zip.

At line 103 in factmngr.c, a patternEntityRecord named factInfo is defined. Before the 4 NULLS in the nested structure, there are two entries: EnvIncrementFactCount followed by EnvDecrementFactCount. These are inthe wrong order. Per the entityRecord declaration in evaluatn.h, the "Decrement" function should appear before the "Increment" function.

This seems like a fairly significant error, so I'm unsure as to why it only seems to manifest itself in a few specific cases -- but correcting this error seems to fix the segmentation faults I was seeing.

Thanks!

Regards,

EnvIncrementFactCount call in clips_fact_assign

wtdoor — Thu, 28 Dec 2017 19:39:56 -0000

Hello --

I've been using PyCLIPS in a project and ran into an issue where CLIPS would occasionally segfault while running. In an attempt to track the issue down, I started running a test case in a loop over and over again and noticed increasingly long run times.

At the beginning of each loop, I'd perform a clips.Reset(), a number of asserts, and a clips.Run(). However, I noticed that the CLIPS call to RemoveGarbageFacts() would take longer and longer with each call -- and it appeared that the linked list of facts maintained by CLIPS continued to get longer and longer.

RemoveGarbageFacts() checks the busyCount of each fact to determine whtehr or not it is still in use before performing cleanup and removing the fact from the list. I noticed that although I had used clips.Reset() and the EnvRetract() had been called on each of the facts, the busyCount was still non-zero as the garbage removal process took place.

It looked to me like the busyCount value was being double incremented when a clips.Assert() call took place. Looking more closely at the code, it appears that when clips.Assert() is used to place a fact into working memory, clips_fact_assign() in clipsmodule.c explicitly calls EnvIncrementFactCount(). This causes busyCount to increment one extra time, such that when my clips.Reset() is executed, the busyCount on each fact I've asserted never goes to 0 and is never cleaned up by the garbage collection.

I hope that makes sense! Thanks!

Regards,

python-call: returning a list of fact refs

Gitttt — Fri, 05 May 2017 08:22:56 -0000

Has this been fixed? I am having the same problems and the code seems not to have changed.

PyCLIPS with Python 3.5 or greater

Mani — Wed, 14 Sep 2016 18:07:04 -0000

Any ideas about bytearrays and unicodes? Anyone ?

PyCLIPS with Python 3.5 or greater

Mani — Wed, 14 Sep 2016 10:34:16 -0000

New dev branch is created on git hub in my repo. I merged your pull request to the dev branch too. Thanks a lot

Is anyone else following this? more tests are welcome . Tim send me his coredump and a backtrace. So far I couldn't reproduce the segfault that he gets on line 259 of test_cycnlst.py

Mani

PyCLIPS with Python 3.5 or greater

Mani — Tue, 13 Sep 2016 14:03:08 -0000

Ok as far as unit tests are concerned every thing is working.

The segfault is still there but it only happens on terminating python , I have a couple of back traces and core dumps that points to the a PyDict_GetItem in bufdict_Remove after its called by python finalize. So its probably a reference counting problem but still not sure where exactly. needs more time in gdb.

Have a look at this
https://github.com/manisabri/pyclips/commit/527d5aaa3d25a6dadb51a90c7029c396edca584d

I'll appreciate if anyone have a go and throws some stuff at it so we find more issues to resolve as not all is covered in unit tests

Thanks
Mani

PyCLIPS with Python 3.5 or greater

Mani — Tue, 13 Sep 2016 11:07:32 -0000

+1 for the dev branch
I'll creat a dev branch and move my commits from master to that after I figured out what's going on in the clips.Eval function

What I'm trying to achive for now is to make it work as intended in python 3.5 with minimal changes. Updating the documents is not my priority for now. But I suppose not much will change except the obvious difference between py 2.7 and 3.5 like map function returning a generator instead of a list and such

At the moment I refactored the test suite so it can run tests separately which make debugging easier

From what I gathered in the past few days PyClips2 branch is inteded to be the better design but its behind on functionality compared to pyclips-1.1_clips-6.30 and it breaks compatibility with pyclips1.

I guess at some point after the master/dev is ported to 3.5 we should consider what we want to do. So far I think out options are:

Study the pyclips2 design and adapt it if we liked it then add the 6.30 functionality
Insert the 6.30 functionality and then adapt the pyclips2 design
Insert the 6.30 functionality and ditch the pyclips2 design ( not so keen on this myself )

It will be very helpful if the original author of the package comes around at some point and give us some insights

What versions of pyclips are you using Tim? On what Platforms?

Mani

PyCLIPS with Python 3.5 or greater

Mani — Mon, 12 Sep 2016 19:30:10 -0000

Here you are
https://github.com/manisabri/pyclips

Master branch is changed ... has the issues I mentioned above.
My rig runs ubuntu 16.04 LTS , but some libs are not from canonical repos , but I guess it shouldn't be any problem

Recent posts to Discussion

Windows Installer Issue

Error in factmngr.c in CLIPSSrc.zip

Error in factmngr.c in CLIPSSrc.zip

EnvIncrementFactCount call in clips_fact_assign

python-call: returning a list of fact refs

PyCLIPS with Python 3.5 or greater

PyCLIPS with Python 3.5 or greater

PyCLIPS with Python 3.5 or greater

PyCLIPS with Python 3.5 or greater

PyCLIPS with Python 3.5 or greater