I wrote a small character based regex engine specifically to support a lexer. However, the last thing lisp needs is another regex package. cl-ppcre is so nice I would like to adopt it instead. However, since scanners try to find the first place in a string that provides a match, it can be impractical to use on very long strings. Is it possible to create a scanner that only matches at the start index?
Matt
On Sun, Sep 28, 2008 at 14:02, Matthew D. Swank akopa.gmane.poster@gmail.com wrote:
Is it possible to create a scanner that only matches at the start index?
I might be confused, but isn't '^' what you need?
-Hans
On Sun, 28 Sep 2008 14:19:21 -0400 "Hans Hübner" hans@huebner.org wrote:
On Sun, Sep 28, 2008 at 14:02, Matthew D. Swank akopa.gmane.poster@gmail.com wrote:
Is it possible to create a scanner that only matches at the start index?
I might be confused, but isn't '^' what you need?
Probably, I just hadn't thought of it. I could pre-pend a '^' to each choice.
Thanks, Matt
On Sun, 28 Sep 2008 14:19:21 -0400 "Hans Hübner" hans@huebner.org wrote:
On Sun, Sep 28, 2008 at 14:02, Matthew D. Swank akopa.gmane.poster@gmail.com wrote:
Is it possible to create a scanner that only matches at the start index?
I might be confused, but isn't '^' what you need?
I tried using a contruct like `(:sequence :start-anchor (:regex ,regex)) where regex is a pcre string, but matching still takes for ever (as in I gave up after 10 min) when slurping a moderately sized file (400k). Note, matching works fine for files under 1k, or if I break it up into lines for line oriented input.
Matt
On Sun, 28 Sep 2008 14:15:40 -0500, "Matthew D. Swank" akopa.gmane.poster@gmail.com wrote:
I tried using a contruct like `(:sequence :start-anchor (:regex ,regex)) where regex is a pcre string, but matching still takes for ever (as in I gave up after 10 min) when slurping a moderately sized file (400k). Note, matching works fine for files under 1k, or if I break it up into lines for line oriented input.
Show us the regex you were using and some test data and then maybe we can help you to optimize it.
I suppose you read this?
http://weitz.de/cl-ppcre/#blabla
Edi.
On Sun, 28 Sep 2008 21:31:05 +0200 Edi Weitz edi@agharta.de wrote:
On Sun, 28 Sep 2008 14:15:40 -0500, "Matthew D. Swank" akopa.gmane.poster@gmail.com wrote:
I tried using a contruct like `(:sequence :start-anchor (:regex ,regex)) where regex is a pcre string, but matching still takes for ever (as in I gave up after 10 min) when slurping a moderately sized file (400k). Note, matching works fine for files under 1k, or if I break it up into lines for line oriented input.
Show us the regex you were using and some test data and then maybe we can help you to optimize it.
I suppose you read this?
http://weitz.de/cl-ppcre/#blabla
Edi. _______________________________________________ cl-ppcre-devel site list cl-ppcre-devel@common-lisp.net http://common-lisp.net/mailman/listinfo/cl-ppcre-devel
Well the regexes are defined in the lexers in this file: http://common-lisp.net/~mswank/apache-ppcre.lisp
The lexer api is in this file: http://common-lisp.net/~mswank/cl-ppcre-lexer.lisp
Finally, the log file I'm lexing: http://lcpug.asternix.com/pub/Main/ApacheLogProject/access.log
Compare (with-open-file (in "access.log") (let ((foo (stream-gen *apache-pcrelex-line* in))) (time (loop :for x := (funcall foo) :unless x :return nil))))
with
(with-open-file (in "access.log") (let ((foo (stream-gen *apache-pcrelex* in))) (time (loop :for x := (funcall foo) :unless x :return nil))))
When I slurp the entire file into a string the matches seem to be taking about a tenth of a second for each token.
Matt
On Sun, 28 Sep 2008 15:28:07 -0500, "Matthew D. Swank" akopa.gmane.poster@gmail.com wrote:
Well the regexes are defined in the lexers in this file: http://common-lisp.net/~mswank/apache-ppcre.lisp
The lexer api is in this file: http://common-lisp.net/~mswank/cl-ppcre-lexer.lisp
Finally, the log file I'm lexing: http://lcpug.asternix.com/pub/Main/ApacheLogProject/access.log
Compare (with-open-file (in "access.log") (let ((foo (stream-gen *apache-pcrelex-line* in))) (time (loop :for x := (funcall foo) :unless x :return nil))))
with
(with-open-file (in "access.log") (let ((foo (stream-gen *apache-pcrelex* in))) (time (loop :for x := (funcall foo) :unless x :return nil))))
When I slurp the entire file into a string the matches seem to be taking about a tenth of a second for each token.
Sorry, I don't have the time to read the entire application right now. Can you boil this down to a single application of PPCRE:SCAN which is too slow?
Thanks, Edi.
Hi Matthew,
You are probably not doing the same thing with the "line oriented approach" and the "full file in one string" approach.
With full file in, if not taking care of stopping the scan at end of each line (if you want a line by line scanning as you suggest by trying such an approach as well), I guess your are scanning until the end of the full string for each line (which for sure is very expensive).
But that's just a guess as I've only had a very quick look to your code :-)
Cheers, Sebastien.
Matthew D. Swank a écrit :
On Sun, 28 Sep 2008 21:31:05 +0200 Edi Weitz edi@agharta.de wrote:
On Sun, 28 Sep 2008 14:15:40 -0500, "Matthew D. Swank" akopa.gmane.poster@gmail.com wrote:
I tried using a contruct like `(:sequence :start-anchor (:regex ,regex)) where regex is a pcre string, but matching still takes for ever (as in I gave up after 10 min) when slurping a moderately sized file (400k). Note, matching works fine for files under 1k, or if I break it up into lines for line oriented input.
Show us the regex you were using and some test data and then maybe we can help you to optimize it.
I suppose you read this?
http://weitz.de/cl-ppcre/#blabla
Edi. _______________________________________________ cl-ppcre-devel site list cl-ppcre-devel@common-lisp.net http://common-lisp.net/mailman/listinfo/cl-ppcre-devel
Well the regexes are defined in the lexers in this file: http://common-lisp.net/~mswank/apache-ppcre.lisp
The lexer api is in this file: http://common-lisp.net/~mswank/cl-ppcre-lexer.lisp
Finally, the log file I'm lexing: http://lcpug.asternix.com/pub/Main/ApacheLogProject/access.log
Compare (with-open-file (in "access.log") (let ((foo (stream-gen *apache-pcrelex-line* in))) (time (loop :for x := (funcall foo) :unless x :return nil))))
with
(with-open-file (in "access.log") (let ((foo (stream-gen *apache-pcrelex* in))) (time (loop :for x := (funcall foo) :unless x :return nil))))
When I slurp the entire file into a string the matches seem to be taking about a tenth of a second for each token.
Matt
On Mon, 29 Sep 2008 13:16:29 +0200 Sébastien Saint-Sevin seb-cl-mailist@matchix.com wrote:
Hi Matthew,
You are probably not doing the same thing with the "line oriented approach" and the "full file in one string" approach.
With full file in, if not taking care of stopping the scan at end of each line (if you want a line by line scanning as you suggest by trying such an approach as well), I guess your are scanning until the end of the full string for each line (which for sure is very expensive).
But that's just a guess as I've only had a very quick look to your code :-)
Cheers, Sebastien.
Well, the lexer code is line agnostic; i.e. you could replace 'end of each line' with any old stop. What it does is adjust the start index as it matches tokens.
One thing I did notice is that I read the file into an adjustable vector, and that is the string I pass to the scanners. I suppose ppcre has to coerce that every time a scanner runs?
Matt
"Matthew D. Swank" akopa.gmane.poster@gmail.com writes:
One thing I did notice is that I read the file into an adjustable vector, and that is the string I pass to the scanners. I suppose ppcre has to coerce that every time a scanner runs?
Yes, scan needs a simple-string. From the SCAN docs:
target-string will be coerced to a simple string if it isn't one already.
Cheers, Chris Dean
On Mon, 29 Sep 2008 10:37:00 -0700 Chris Dean ctdean@sokitomi.com wrote:
"Matthew D. Swank" akopa.gmane.poster@gmail.com writes:
One thing I did notice is that I read the file into an adjustable vector, and that is the string I pass to the scanners. I suppose ppcre has to coerce that every time a scanner runs?
Yes, scan needs a simple-string. From the SCAN docs:
target-string will be coerced to a simple string if it isn't one already.
Coercing the slurped file to a simple string makes things work swimmingly.
Thanks,
Matt
cl-ppcre-devel@common-lisp.net
-
Chris Dean -
Edi Weitz -
Hans Hübner -
Matthew D. Swank -
Sébastien Saint-Sevin