Is there a maximum length of a regular expression in python?

[olekristianvillabo]

I have a regular expression that is approximately 100k bytes. (It is
basically a list of all known norwegian postal numbers and the
corresponding place with | in between. I know this is not the intended
use for regular expressions, but it should nonetheless work.

the pattern is
ur'(N-|NO-)?(5259 HJELLESTAD|4026 STAVANGER|4027 STAVANGER........|8305
SVOLVÆR)'

The error message I get is:
RuntimeError: internal error in regular expression engine

 

[Steve Holden]

I have a regular expression that is approximately 100k bytes. (It is
basically a list of all known norwegian postal numbers and the
corresponding place with | in between. I know this is not the intended
use for regular expressions, but it should nonetheless work.

the pattern is
ur'(N-|NO-)?(5259 HJELLESTAD|4026 STAVANGER|4027 STAVANGER........|8305
SVOLVÆR)'

The error message I get is:
RuntimeError: internal error in regular expression engine

And I'm not the least bit surprised. Your code is brittle (i.e. likely
to break) and cannot, for example, cope with multiple spaces between the
number and the word(s). Quite apart from breaking the interpreter

I'd say your test was the clearest possible demonstration that there
*is* a limit.

Wouldn't it be better to have a dict keyed on the number and containing
the word (which you can construct from the same source you constructed
your horrendously long regexp)?

Then if you find something matching the pattern (untested)

ur'(N-|NO-)?((\d\d\d\d)\s*([A-Za-z ]+))'

or something like it that actually works (I invariably get regexps wrong
at least three times before I get them right) you can use the dict to
validate the number and name.

Quite apart from anything else, if the text line you are examining
doesn't have the right syntactic form then you are going to test
hundreds of options, none of which can possibly match. So matching the
syntax and then validating the data identified seems like a much more
sensible option (to me, at least).

regards
Steve

 

[Fredrik Lundh]

I have a regular expression that is approximately 100k bytes. (It is
basically a list of all known norwegian postal numbers and the
corresponding place with | in between. I know this is not the intended
use for regular expressions, but it should nonetheless work.

the pattern is
ur'(N-|NO-)?(5259 HJELLESTAD|4026 STAVANGER|4027 STAVANGER........|8305
SVOLVÆR)'

The error message I get is:
RuntimeError: internal error in regular expression engine

you're most likely exceeding the allowed code size (usually 64k).

however, putting all postal numbers in a single RE is a horrid abuse of the RE
engine. why not just scan for "(N-|NO-)?(\d+)" and use a dictionary to check
if you have a valid match?

postcodes = {
"5269": "HJELLESTAD",
...
"9999": "ØSTRE FJORDVIDDA",
}

for m in re.finditer("(N-|NO-)?(\d+) ", text):
prefix, number = m.groups()
try:
place = postcodes[number]
except KeyError:
continue
if not text.startswith(place, m.end()):
continue
# got a match!
print prefix, number, place

</F>

 

[Roy Smith]

I have a regular expression that is approximately 100k bytes. (It is
basically a list of all known norwegian postal numbers and the
corresponding place with | in between. I know this is not the intended
use for regular expressions, but it should nonetheless work.

the pattern is
ur'(N-|NO-)?(5259 HJELLESTAD|4026 STAVANGER|4027 STAVANGER........|8305
SVOLVÆR)'

The error message I get is:
RuntimeError: internal error in regular expression engine

I don't know of any stated maximum length, but I'm not at all surprised
this causes the regex compiler to blow up. This is clearly a case of regex
being the wrong tool for the job.

I'm guessing a dictionary, with the numeric codes as keys and the city
names as values (or perhaps the other way around) is what you want.

 

[Frithiof Andreas Jensen]

I have a regular expression that is approximately 100k bytes. (It is
basically a list of all known norwegian postal numbers and the
corresponding place with | in between. I know this is not the intended
use for regular expressions, but it should nonetheless work.

Err. No.

A while back it was established in this forum that re's per design can have
a maximum of 99 match groups ... I suspect that every "|" silently consumes
one match group.

 

[Fredrik Lundh]

Err. No.

A while back it was established in this forum that re's per design can have
a maximum of 99 match groups ... I suspect that every "|" silently consumes
one match group.

nope. this is a code size limit, not a group count limit.

</F>

 

[Bryan Olson]

I don't know of any stated maximum length, but I'm not at all surprised
this causes the regex compiler to blow up. This is clearly a case of regex
being the wrong tool for the job.

Does no one care about an internal error in the regular expression
engine?

 

[Roy Smith]

Does no one care about an internal error in the regular expression
engine?

I think the most that could be said here is that it should probably produce
a better error message.

 

[Steve Holden]

Does no one care about an internal error in the regular expression
engine?

Not one that requires parsing a 100 kilobyte re that should be replaced
by something more sensible, no.

regards
Steve

 

[Paul Rubin]

Not one that requires parsing a 100 kilobyte re that should be
replaced by something more sensible, no.

If the internal error means the re engine bumped into some internal
limit and gracefully raised an exception, then fine. If "internal
error" means the re engine unexpectedly got into some inconsistent
internal state, then threw up its hands and barfed after discovering
the error sometime later, that's bad. Does nobody care which it is?

 

[Roy Smith]

If the internal error means the re engine bumped into some internal
limit and gracefully raised an exception, then fine. If "internal
error" means the re engine unexpectedly got into some inconsistent
internal state, then threw up its hands and barfed after discovering
the error sometime later, that's bad. Does nobody care which it is?

The nice thing about an open source project is that if nobody else gets
excited about some particular issue which is bothering you, you can take a
look yourself.

(from Python-2.3.4/Modules/_sre.c):

static void
pattern_error(int status)
{
switch (status) {
case SRE_ERROR_RECURSION_LIMIT:
PyErr_SetString(
PyExc_RuntimeError,
"maximum recursion limit exceeded"
);
break;
case SRE_ERROR_MEMORY:
PyErr_NoMemory();
break;
default:
/* other error codes indicate compiler/engine bugs */
PyErr_SetString(
PyExc_RuntimeError,
"internal error in regular expression engine"
);
}
}

I suppose one man's graceful exit is another man's barf.

 

[Tim Peters]

[Bryan Olson]
[Steve Holden]

Not one that requires parsing a 100 kilobyte re that should be replaced
by something more sensible, no.

I care: this is a case of not detecting information loss due to
unchecked downcasting in C, and it was pure luck that it resulted in
an internal re error rather than, say, a wrong result. God only knows
what other pathologies the re engine could tricked into exhibiting
this way. Python 2.5 will raise an exception instead, during regexp
compilation (I just checked in code for this on the trunk; with some
luck, someone will backport that to 2.4 too).

 

[Steve Holden]

[Bryan Olson]


[Steve Holden]

Not one that requires parsing a 100 kilobyte re that should be replaced
by something more sensible, no.

I care: this is a case of not detecting information loss due to
unchecked downcasting in C, and it was pure luck that it resulted in
an internal re error rather than, say, a wrong result. God only knows
what other pathologies the re engine could tricked into exhibiting
this way. Python 2.5 will raise an exception instead, during regexp
compilation (I just checked in code for this on the trunk; with some
luck, someone will backport that to 2.4 too).

Just goes to show you, ignorance is bliss.
What would we do without you, Tim?

regards
Steve

 

[Frithiof Andreas Jensen]

Roy Smith wrote:

Does no one care about an internal error in the regular expression
engine?

Yes, but - given the example - In about the same way that I care about an
internal error in my car engine after dropping a spanner into it ;-)

 

[fuzzylollipop]

this should really be posted to http://www.thedailywtf.com/, I wonder
if they have a german version of TheDailyWTF.com?