Support custom regex

[yocontra]

It would be cool if this supported passing in a regex instead of a string that becomes a regex.

[sindresorhus]

👍

[sandinmyjoints]

👍

[nicolashery]

👍

[malgorithms]

I imagine the difficulty here is the cross-boundary matching... @eugeneware - is that right?

If Eugene's constructing a regular expression out of a string you pass, he only needs to keep strlen(search term) data around from the previous chunk to watch for replacements.

I would like this too...just saying I wouldn't know how it would be implemented without using a non-native regular expression implementation.

I suppose it could support regular expressions but also require you pass it a maximum match length at the same time.

// remove html comments up to 1024 chars
// longer comments would not get replaced
.pipe(replaceStream(/<!--(.|\s)*-->/g, '', { max_match_len: 1024 } ))

[yocontra]

@malgorithms max_match_len requirement on custom regexs sounds fine to me, that would solve most cases

[eugeneware]

Correct, you could end up buffering the entire stream in memory if you
don't get a hit, which would kind of suck.

But implementing a maximum match length might be an option, though not
deterministic of course.

On Fri, Apr 25, 2014 at 7:37 AM, Eric Schoffstall
notifications@github.comwrote:

@malgorithms https://github.com/malgorithms max_match_len requirement
on custom regexs sounds fine to me, that would solve most cases

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/9#issuecomment-41335792
.

Eugene Ware
Chief Executive Officer

Phone: +61 3 9955 7041
Email: eugene@noblesamurai.com
Twitter: @eugeneware http://twitter.com/EugeneWare

Noble Samurai Pty Ltd
Level 1, 234 Whitehorse Rd
Nunawading, Victoria, 3131, Australia

noblesamurai.com http://www.noblesamurai.com/ | eugeneware.com |
facebook.com/Eugene.S.Ware http://www.facebook.com/Eugene.S.Ware

[malgorithms]

@eugeneware - what do you mean "not deterministic" in this context? That it would be unpredictable based on boundaries?

Like, for a simple example, if you decided you wanted to replace /a+/ with just b then if you happened to just swallow a few a's but then hit a boundary, you might mistakenly replace it with a b even though there were more a's coming, which should've been part of the match, because it was all less than max_match_len?

I think if you kept a moving window of data of size (max_match_length * 2) and only performed replacements on matches which began in the first half of the window, there wouldn't be any ambiguities. I think. Does that sound right to you? It would swallow up to max_match_length of repeating a's and replace them with a b, as expected by the call.

[eugeneware]

The moving window approach would definitely work. Should be easy enough to
implement. Happy to take a PR if you'd like to take a shot at this too :-)

On Fri, Apr 25, 2014 at 11:34 PM, Chris Coyne notifications@github.comwrote:

@eugeneware https://github.com/eugeneware - what do you mean "not
deterministic" in this context? That it would be unpredictable based on
boundaries?

Like, for a simple example, if you decided you wanted to replace /a+/with just
b then if you happened to just swallow a few a's but then hit a boundary,
you might mistakenly replace it with a b even though there were more a's
coming, which should've been part of the match, because it was all less
than max_match_len?

I think if you kept a moving window of data of size (max_match_length * 2)
and only performed replacements on matches which began in the first half of
the window, there wouldn't be any ambiguities. I think. Does that sound
right to you? It would swallow up to max_match_length of repeating a's
and replace them with a b, as expected by the call.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/9#issuecomment-41392722
.

Eugene Ware
Chief Executive Officer

Phone: +61 3 9955 7041
Email: eugene@noblesamurai.com
Twitter: @eugeneware http://twitter.com/EugeneWare

Noble Samurai Pty Ltd
Level 1, 234 Whitehorse Rd
Nunawading, Victoria, 3131, Australia

noblesamurai.com http://www.noblesamurai.com/ | eugeneware.com |
facebook.com/Eugene.S.Ware http://www.facebook.com/Eugene.S.Ware

[malgorithms]

Fair enough! I am unlikely to do this as I no longer need it, but if I get some hobby time...

[mehtaphysical]

I can take this on. I just used this in a project using regex. Here is the changes I made: https://github.com/mehtaphysical/replacestream/tree/add-regex

I'll make a pull request so we can create a dialog about it. But first I want to make a test.

[eugeneware]

Thanks @mehtaphysical that would be great!