Nicolas Lehuen's Weblog - Latest Comments

Re: Pure bliss with MongoDB

Sougata Pal — Sat, 31 Dec 2011 05:52:17 -0000

Nice article. There is php implementation of the above sharing available named as MongoLantern

Re: pytst 1.18

Nicolas Lehuen — Mon, 12 Sep 2011 16:07:48 -0000

Hi dbg,

The current scan implementation is greedy, it find longest matches and "consume" them. This is precisely what I needed for my projects (I even have a version of the scan algorithm that only accept matches on stop word boundaries), but of course it may not be what you need.

What you need is an implementation of the Aho-Corasick algorithm (http://en.wikipedia.org/wiki/A.... Initially I tried to implement a custom-made, greedy, non-backtracking algorithm that I thought was Aho-Corasick. But obviously, being greedy, it could not be A-C !

Then I found the bug the blog entry mentioned, and I dumped the whole thing for a simpler, safer, greedy-but-backtracking algorithm. Truth is, I don't use pytst anymore and I have no more time to maintain it. But I could not just let the project out there with a bug in the scan algorithm, so I did this last release.

The whole source code is under github (https://github.com/nlehuen/pyt... so don't hesitate to fork the project if needed.

Regards,
Nicolas

Re: pytst 1.18

dbg — Mon, 12 Sep 2011 15:09:18 -0000

Here is some exact code:

>>> t = tst.TST()>>> t['111'] = 'first'>>> t['222'] = 'second'>>> t['112'] = 'third'>>> string = '111222111'

>>> class tst_findall(object):... def __init__(self):... self.pos = 0... self.scores = []... self.locations = []... self.nmers = []... def hit(self,key,length,obj):... if length > 0:... self.scores.append(obj)... self.nmers.append(key)... self.locations.append((self.pos,self.pos+length))... self.pos += abs(length)... def result(self):... return (self.nmers,self.locations,self.scores)

>>> tstf = tst_findall()>>> call_tst_findall = tst.CallableAction(tstf.hit,tstf.result)

>>> t.scan(string,call_tst_findall)(['111', '222', '111'], [(0, 3), (3, 6), (6, 9)], ['first', 'second', 'first'])

I would expect:

>>> t.scan(string,call_tst_findall)(['111','112','222', '111'], [(0, 3), (1, 4), (3, 6), (6, 9)], ['first', 'third', 'second', 'first'])

Re: pytst 1.18

dbg — Mon, 12 Sep 2011 14:59:02 -0000

Hi Nicolas,

Thanks for this useful module! I was looking everywhere for a fast substring search implementation. However, I also seem to still have a problem with the scan function. For instance, if I have a dictionary like {'111','222','112'} and a string '111222111', I would expect it to find '111' at positions 0 and 6, '222' at position 3, and '112' at position 1. Currently it is not finding 112 at all, probably because that match begins inside the match '111'. Is this the intended behavior, or was the 1.18 update supposed to fix this?

Many thanks for your help, and your useful module.

Re: pytst 1.18

Nicolas Lehuen — Tue, 03 May 2011 03:56:15 -0000

Hi Nikolai,

Can you send me your test case (nicolas at lehuen dot com) ? It is very simple so it should work straight away... Unless I've missed something.

Regards,
Nicolas

Re: pytst 1.18

Nikolai Yakovenko — Wed, 27 Apr 2011 18:19:52 -0000

Hi Nicolas,

Do you have a version of PyTST 1.18 that I can download for Mac OS? I downloaded latest ZIP in the link you sent me... but I think it still has this problem.

I am matching words in text preprocessed to make explicit word boundary. So I match strings like " dog " and " cat ". If a text has " dog cat ", then the second won't match. This is the problem that you fixed, right?

I used Aho-Corasick extensively when I worked for Google :-). Trying to use it now too, on an independent project.

Thanks,
Nikolai

Re: SOAP Web Services : awkward. Python and Web Services : painful.

Jeshurun — Tue, 03 Aug 2010 11:21:42 -0000

You have written this in 2006. I'm trying it in 2010 and its still painful!

Re: Pure bliss with MongoDB

Mike Dirolf — Thu, 15 Apr 2010 16:31:03 -0000

Interesting read - thanks for sharing!

Re: A case against Twitter

Fab — Tue, 30 Mar 2010 05:09:44 -0000

Ouais t'as raison, twitter c'est de la merde !

Re: std::vector is very slow ? - Nicolas Lehuen's Weblog

fd — Sat, 13 Feb 2010 06:13:48 -0000

hey bro my std::vector copy takes seven milliseconds in debug mode, but <400 nanoseconds in release mode. It does a LOT of debug stuff.

Re: Mapping my personal web - Nicolas Lehuen's Weblog

Fil — Wed, 09 Dec 2009 15:29:15 -0000

I have a trick. I subscribe to Nicolas Lehuen's Shared Items on Google Reader, and that's well enough interesting stuff to read. Thanks. :-)

Re: Une idée lumineuse ! - Nicolas Lehuen's Weblog

maxi69 — Wed, 30 Sep 2009 11:52:24 -0000

On a un Nico écolo, ;o) peut-être même qu'il finira par s'inscrire chez les Verts ?

Re: A case against Twitter

Nicolas Lehuen — Wed, 12 Aug 2009 12:47:16 -0000

About infrastructure and money : for what it's worth, compare the $55M that Twitter raised according to Techcrunch ( http://www.crunchbase.com/company/t... ) to the $5M FriendFeed raised (http://www.crunchbase.com/company/f...).

Re: A case against Twitter

Nicolas Lehuen — Wed, 12 Aug 2009 10:13:39 -0000

Two more points :

1) SPAM BOTS ! The open broadcast model of Twitter is a true spam magnet and it shows http://www.readwriteweb.com/archive... .

1) Twitter sues a developer using their API for improper (spam related) use : http://www.mytwitterbutler.com/I'm_... . Of course the use may be improper, I don't know of this particular service. But it's ironic to see Twitter turn against developers since they owe them so much.

Re: FriendFeed might actually benefit from Google Wave

Nicolas Lehuen — Fri, 29 May 2009 05:16:15 -0000

Zoho announced that they plan to embrace Google Wave : http://blogs.zoho.com/general/surfi... . This is a good example of the strategy FriendFeed could follow, though of course there is a less apparent functional clash between Google Wave and Zoho...

Re: pytst 1.17

Paul Harrington — Wed, 01 Apr 2009 19:17:09 -0000

I am running the unit tests right now. Thanks very much for your work.

Re: pytst 1.15

yatsu — Thu, 03 Apr 2008 23:40:45 -0000

Hi Nicolas,

It is a really good news for me that you are developing ctst and the Ruby bindings for it.

I don't need to modify Python dependent code to port it to Ruby anymore.

I hope ctst will be released anytime soon.

Re: A tribute to Groovy

Guillaume Laforge — Thu, 21 Feb 2008 04:01:48 -0000

Nicolas, thanks a lot for this great tribute!

Re: A tribute to Groovy

Fabien Gandon — Mon, 18 Feb 2008 03:22:42 -0000

I knew the beginning of the saga. I now know the end.

Knowing you, I just wondered: how long before you find an "even better" framework ?

Just teasing ;-)

As for the applet vs. flash, I would love to see both disappear and be replaced by some powerful compound document profiles http://www.w3.org/2004/CDF/ for the sake of customization, device/context adaptation and accessibility.

Cheers.

Re: PullXML renamed to PushXML, hosted on Google Code

Nicolas Lehuen — Thu, 14 Jun 2007 10:26:04 -0000

Well, I've got not plan to support XPath for now. The purpose of the object model that PushXML builds is precisely to remove the need for XPath in most of the case : instead of /foo[1]/bar[1] you just write $root->foo[0]->bar[0] and you're done. Of course, this doesn't cover the cases where you want to write //bar or even //bar[@type='anyway'], but this is not something I need right now.

I guess the best thing to do would be to implement an XPath library that would know how to navigate simple object trees, as it would be useful even in over situations. A few years ago, I wrote something similar in Java, unfortunately my time is too scarce right now for this kind of little pleasures :).

Re: PullXML renamed to PushXML, hosted on Google Code

marina — Mon, 04 Jun 2007 05:04:17 -0000

Hello,

PushXML looks really nice, but it looks like only simple path like "/foo/bar" can be used. Do you plan to add support for more complex xpath queries ? I currently use simpleXML and I find that xpath support is really essential in my project !

Re: Strings mess

Samrat — Thu, 22 Mar 2007 07:23:27 -0000

Hi,

I am a VC programmer and am grappling with this same issue...

spending hours and days altogether to port legacy classes to newer framework..and the amount of time i spend to solve all the string mess...95% of the entire time...

i have gone wild hunting for the reason for such a huge mess and hence landed on ur blog...

Re: Ma très chère Free.

mme champarou — Mon, 19 Mar 2007 14:48:38 -0000

ah ah ah

j'ai beaucoup aimé

croustillant

Re: Some fun with the Nintendo DS

Azlan — Thu, 15 Mar 2007 03:50:52 -0000

amzingly intriguing , i just hapened appon your root files for the mandelbrot while looking for random homebrew nds files. interesting concept and it accualy works on a real ds

Re: Switched to DotClear 2 beta 6

Nicolas Lehuen — Sun, 11 Mar 2007 10:26:44 -0000

Of course, rewriting the permalink inside my blog so that they point to the new style is not sufficient, because some people out there are pointing to my blog and I can't rewrite their permalinks. So I did a little bit more of mod_rewrite trickery, and everything should be just fine right now.