Author timehorse
Recipients amaury.forgeotdarc, effbot, giampaolo.rodola, gregory.p.smith, mrabarnett, pitrou, terry.reedy, timehorse
Date 2008-09-24.12:06:40
SpamBayes Score 0.0
Marked as misclassified No
Message-id <>

I am really happy that you are making such progress on your engine, but
can I PLEASE ask you to slow down for a moment?  We have a lot of issues
already listed in issue 2636 that is a catch-all for any Python 2.7
Regexp improvements, including your new engine, and I have been working
frantically to try and document all the changes YOU are making here into
the general Regexp 2.7 modification thread and setting up development
trees in my Bazaar VCS repository for your work.

There is also a recommended process for patching which makes it easier
for the moderators to accept your patches which is to provide
dis-entangled functionality and letting each improvement stand on its
own two feet.  In other words, let your engine stand ONLY on it's 2x
speed improvements.  We already have an implementation of Atomic
Grouping / Possessive Qualifiers in issue 2636 but you have a version of
your engine with both.  We have no such 'feature-only' implementation
for Variable-Length Look-Behind, for a Reverse flag, for Positionally
Dependent modifier flags or modifier negation flags, as well as the
zero-width Regular Expression split feature, though you and I completely
agree these would all be great things to have!  The more features you
add to your engine as an all-or-nothing proposition, the less likely the
moderators are going to be to adapt it because it's harder for them to
examine the merits of each individual piece.  That is why issue 2636 is
broken up into items (currently 1 - 18, with your proposals bringing
that up toward 22) and where alternate, combined features are provided
if implementing 1 features would affect the implementation of another.

Please understand that I personally have no problem with you redesigning
large swaths of the Python Regular Expression engine.  I would
personally, like to see one of the design goals of any new engine not
only be speed but better source comments because my main beef with the
current engine is that it took me a month to understand and part of my
redesign in issue 2636 9-1 was to add copious comments to the engine so
that future developers would understand what was going on and be able to
pick up from my work.  I am not proposing we use my 9-1 engine because
it is 8% slower than the current engine and I don't intend to propose
anything slower.  But it would be nice if you could add lots of comments
to your engine so that others could help develop features against it. 
None the less, I will fully support your engine if it does indeed
perform substantially and measurably faster and am happy to see all the
Regexp issues you are finding are finally being implemented, all be it
entangled with your engine.  But let's return to the fundamentals of
what you propose IN THIS THREAD, which simply to propose a new Regexp
Engine which is 2x faster than the existing engine (Which I have
allocated item 9-2 in the issue 2636 thread).  I am not trying to put
more work on your hands -- in fact, what I am trying to do is get us to
co-operate on a better python Regexp Engine so that I can help you to
achieve your goals.  Please read issue 2636 and join the discussion
there; feel free to add any new items you feel are missing from my
existing list.  And remember, each new feature needs tests and
documentation changes.  I have been doing each for any feature I
undertake and would be happy to share those skills with you.  

Let's work together to see your engine be the new model, okay?

Date User Action Args
2008-09-24 12:06:44timehorsesetrecipients: + timehorse, effbot, terry.reedy, gregory.p.smith, amaury.forgeotdarc, pitrou, giampaolo.rodola, mrabarnett
2008-09-24 12:06:43timehorsesetmessageid: <>
2008-09-24 12:06:42timehorselinkissue3825 messages
2008-09-24 12:06:40timehorsecreate