Indefinite length lookbehind

[philja]

I'm always interested in expanding my use of regex so I tried both Doitsu's test case and Kevin's test case and couldn't get either to work in Sigil 2.4.2 on my MacBook.

Doitsu's original Find

(?<!”\s\w+\s)said\. “

and

Kevin's variation :

(?<!”\s\w{1,6}\s)said\. “

both give 'Invalid Regex: offset:0' when I hover on them as does Kevin's other test case

(?<=0x{1,6})y

Presumably, these should work in 2.4.2 on Mac?

Edit:
I must be seriously missing something. I can't get any of these 3 regex Finds to work on my linux or W10 boxes either.

Hovering the cursor over the find term gives exactly the same message each time: 'Invalid Regex: offset:0' which doesn't help me at all.

I can get both

(?<!"\sJack\s)said\. "   and   (?<!"\s\w{4}\s)said\. "

to work with Diatsu's test epub ie it ignores his first line and finds the other two. But as soon as I mess around trying to use a range {1,6} or similar, the regex is described as invalid.

[KevinH]

Yes all 3 systems are using PCRE2 10.44 or earlier which have the bug. The next Sigil will use PCRE2 10.45.

On Linux, you should be easily able to update just the PCRE2 library to version 10.45 and Sigil 2.4.2 should work just fine with these tests then.

[DiapDealer]

Can confirm that the Arch repo version of Sigil 2.4.2 doesn't have the bug--as it's built to use the system version of pcre2 (which is 10.45). If you build your own Sigil 2.4.2 with bundled libs (the default) the bug will manifest.

[philja]

Originally Posted by KevinH

Yes all 3 systems are using PCRE2 10.44 or earlier which have the bug. The next Sigil will use PCRE2 10.45.

On Linux, you should be easily able to update just the PCRE2 library to version 10.45, and Sigil 2.4.2 should work just fine with these tests then.

Just checked in my UbuntuStudio 24.04.02 LTS and all pcre2 packages in the repo seem to be 10.42, so it looks like waiting for the next version of Sigil.

[DiapDealer]

You built your own version of Sigil on 24.04, (I don't believe sigil 2.4.2 is available for 24.04), right? If so, there's no need to wait for anything. Just build Sigil from the latest GitHub master source and don't configure it to use the system libs. That way it will be built with the pcre2 10.45 now bundled with Sigil.

Ubuntu 24.04 won't likely be upgrading their version of pcre2, so you're going to need to remember to not use -DUSE_SYSTEM_LIBS=1 when configuring Sigil--whenever you build Sigil.

[ElMiko]

Just a pointless sidebar that was probably addressed 5 years ago with the release of 1.0, thank you for adding the Regex validation into the program, and making it sensitive not to external regex rules but to Sigil's compatibility. Otherwise we'd have been spinning our wheels a lot longer even after landing on nominally valid regex like:

(?<!”\s\w{1,6}\s)said\. “

Although my decade-overdue upgrade to 2.4.2 seems to have produced a lot of panicked screeching from me in the last couple of weeks, I swear 1) I'm not as panicked as it may seem, and 2) I'm increasingly sympathetic to the view that once you go 2.4.2, you can't go back.

[Doitsu]

Originally Posted by philja

I'm always interested in expanding my use of regex so I tried both Doitsu's test case [...]

Both you and ElMiko apparently missed the following:

• I wrote a simple Sigil regex search plugin
• That plugin supports a third-party Python regex library called "regex"
• The regex Python library supports variable-length lookbehínds
• I.e. the plugin could be used as a stop-gap measure.
  (It'll search all files selected in the book browser and display the matches in the validation´window.)

[philja]

Originally Posted by DiapDealer

You built your own version of Sigil on 24.04, (I don't believe sigil 2.4.2 is available for 24.04), right? If so, there's no need to wait for anything. Just build Sigil from the latest GitHub master source and don't configure it to use the system libs. That way it will be built with the pcre2 10.45 now bundled with Sigil.

Ubuntu 24.04 won't likely be upgrading their version of pcre2, so you're going to need to remember to not use -DUSE_SYSTEM_LIBS=1 when configuring Sigil--whenever you build Sigil.

You're so persuasive !! I was so far down this rabbit hole that, after saying Howdo! to Alice, I took up your suggestion and rebuilt Sigil from the latest Master.

The regex problem has now gone on my Linux box - (and I even remembered that testplugin has to used on Sigil's default empty file!!).

Thanks for the push.

[philja]

Originally Posted by Doitsu

Both you and ElMiko apparently missed the following:

• I wrote a simple Sigil regex search plugin
• That plugin supports a third-party Python regex library called "regex"
• The regex Python library supports variable-length lookbehínds
• I.e. the plugin could be used as a stop-gap measure.
  (It'll search all files selected in the book browser and display the matches in the validation´window.)

Not missed. I just got more hooked up to rebuilding on Linux. I'll follow up with your suggestion for my other machines. Thanks.

[DiapDealer]

Originally Posted by philja

You're so persuasive !! I was so far down this rabbit hole that, after saying Howdo! to Alice, I took up your suggestion and rebuilt Sigil from the latest Master.

The regex problem has now gone on my Linux box - (and I even remembered that testplugin has to used on Sigil's default empty file!!).

Thanks for the push.

Practice does everybody good!

For all intents and purposes, you now have what will be the upcoming Sigil 2.5.0 release. So unless something comes up this week, you won't need to rebuild after the upcoming release.

[philja]

Originally Posted by Doitsu

Both you and ElMiko apparently missed the following:

• I wrote a simple Sigil regex search plugin
• That plugin supports a third-party Python regex library called "regex"
• The regex Python library supports variable-length lookbehínds
• I.e. the plugin could be used as a stop-gap measure.
  (It'll search all files selected in the book browser and display the matches in the validation´window.)

I tried your plugin on my MacBook with success against the tests specified earlier in this thread.

My main problem was watching plugin runner's blue button oscillating backwards and forwards for several minutes before I remembered that you said something about having to select the regex engine.

I couldn't find where to make that selection. I finally tracked down a postage-stamp-sized GUI which had opened behind the main Sigil window.

Once I entered the regex in that gui, it worked well.

I've tried several times on my MacBook, and that little gui always hides away behind the main window.

[DiapDealer]

HighDPI scaling on Linux and macOS retina scaling do terrible things to tkinter widgets. Not only are they often tiny, but they'll only match the system light/dark theme on macOS.

[ElMiko]

@Doitsu - Sorry 'bout that! You're right; i misunderstood. I thought your plugin was a straight regex validation tool. I didn't understand from your post that it was also a functional search/replace module.