Affinity Publisher: Union of Regex routines

[Pyanepsion]

Hello everyone,

I would like to create a Regex routine that replaces the offending spaces with a narrow no-break space <nbsp> (U+202F) in one pass in one routine instead of two.

Routine 1. Replace all erroneous spaces before colon (:), right-pointing double angle quotation mark (»), percent sign (%) with a narrow no-break space

Search for:
\s(?=(:|»|%))
Replacement:
<nnbsp>

Routine 2. Replace erroneous spaces after left-pointing double angle quotation mark («) with a narrow no-break space <NNBSP> (U+202F).

Search for:
(«)\s
Replacement:
\1<nnbsp>

Is this possible?

Is it possible to insert a unicode value into Regex formulas such as \u+202f, or to indicate a variable such as <nnbsp> that appears in the occurrence display?

Or, is it necessary to paste a narrow non-breaking space <nnbsp> (obtained by U+202F from the text)?

Thank you for your explanations.

regex.afpub

Edited February 6, 2023 by Pyanepsion
Correction of the English translation name error on («) and (»).

[walt.farrell]

2 hours ago, Pyanepsion said:

left-pointing double angle quotation mark (»)

I would quibble that that is a right-pointing mark (though perhaps it appears at the left of a string that is quoted?) Similarly comment for your other mark, which "points" left.

But that's not relevant to the question.  

There are very few regular-expression implementations that are documented as supporting conditional processing in the replacement value, and Boost (used in Publisher) is not one of them. So I don't think you can do what you want, but I can't try it to confirm until later.

[Pyanepsion]

19 hours ago, walt.farrell said:

I would quibble that that is a right-pointing mark (though perhaps it appears at the left of a string that is quoted?)

Yes, I had reversed the names of left-pointing double angle quotation mark («) and right-pointing double angle quotation mark (») in the English translation. That is corrected.

The important thing here is:

• either to find a solution for the points raised in search and replace:
  • combine several Regex actions into one,
  • use \u+ for Unicode variables such as \u+202f,
  • use variables for constants such as <nnbsp>,
  • be able to save and load Regex scripts).
• or suggest features to improve the use of Regex in the Affinity suite.

[laurent32]

23 hours ago, Pyanepsion said:

Search for:
\s(?=(:|»|%))
Replacement:
<nnbsp>

I'm not a regex genius but I try to understand what I read…

Shouldn't we have <nnbsp>\1 ? or is it implicit ?

[Andreas CH]

 

Hier der Ausdruck für das linke Guillemet

(?<=«)\s

und hier der Ausdruck für alle Zeichen

\s(?=(:|»|%))|(?<=«)\s

Leider kann man das <NNBSP> nur aus dem Layout in das Ersetzen kopieren. 😞

 

 

Edited February 6, 2023 by Andreas CH

[Pyanepsion]

@laurent32, Here is a step-by-step explanation to help understand this regex formula.

Search for:
\s(?=(:|»|%))
Replacement:
<nnbsp>

The regex formula is used to search and replace characters in a text string. Here is a step-by-step explanation of this formula:

• Search for:
  1. \s stands for ‘space’. It looks for a space in the text.
  2. (?=(:|"|%)) is a ‘positive look ahead’. It looks at what follows the space and checks if it is one of the characters (:), (;), (%).
  3. If a space is found and the following characters match one of the specified characters, the replacement of this space will be performed.
• Replacement:
  1. <nnbsp> defines the replacement that will be performed. In this case, the space will be replaced with the character ‘narrow no-break space’ (<nnbsp>).

End result:

The sentence ‘Il lui dit : « je sais. C’est 15 %. ».’ will be changed to ‘Il lui dit<nnbsp>: « je sais. C’est 15<nnbsp>%.<nnbsp>».’.

[Alfred]

21 minutes ago, Pyanepsion said:

(?=(:|"|%)) is a ‘positive look ahead’. It looks at what follows the space and checks if it is one of the characters (:), (;), (%).

For the sake of clarity, I believe you meant to type:

Quote

(?=(:|»|%)) is a ‘positive look ahead’. It looks at what follows the space and checks if it is one of the characters (:), (»), (%).

 

[laurent32]

1 hour ago, laurent32 said:

Shouldn't we have <nnbsp>\1 ? or is it implicit ?

My question here was :

Since we use parentheses…

\s(?=(:|»|%))

Shouldn't we have <nnbsp>\1 for the replace part ?

I thought \1 referred to what is between () ?

But, maybe because we have only one () we don't need to say \1 ?

[laurent32]

43 minutes ago, Andreas CH said:

\s(?=(:|»|%))|(?<=«)\s

That seems quite interesting… like having an "or" between 2 lookups…

Looks like what you're looking for @Pyanepsion ?

[walt.farrell]

49 minutes ago, laurent32 said:

Shouldn't we have <nnbsp>\1 for the replace part ?

No, as this particular use of () does not indicate a capture group. It indicates, because of the added the ?=,  that it is instead a positive look-ahead group. You would only use a replacement like \1 with a capture group.

[walt.farrell]

@Pyanepsion's request for a union of the two expressions can be accomplished, in some regular expression engines. It requires that they support a replacement function called conditional replacement, in which the replacement string can handle a kind of if-then-else processing. If conditional replacement is supported, the syntax is different for it among the different engines, so I can't show a definite syntax for it. But in one possible form it looks something like: ?1true:false

In that replacement syntax:

• ?1 signifies a conditional replacement, to be applied based on whether the search/match process set capture group #1. 
• The "true" text is the replacement if capture group 1 was set.
• The : separates the true/false values
• the "false" text is the replacement if capture group 1 was not set.
   

Thus, you can have a search like (y)|n and a replacement like ?\1yes:no and if you apply it to the string yyn the result would be yesyesno

After some research, I have found that:

• The Boost regex engine says that it has support for conditional replacement.
• However, the application (e.g., Publisher) that is using Boost must specifically enable the conditional replacement support when it calls the replace function.

So: It might work, if Serif coded Publisher to enable the conditional replacement.

However, when I tried both of the forms of conditional replacement that I could find documented for Boost, neither worked. This should mean that either:

• Publisher does not enable the support for conditional replacement; or
• I have not found the right format for the conditional replacement function in Boost; or
• This rarely used function is broken in Boost.

My research continues.

But if anyone from Serif is reading this, and can check with the developers whether the format_all option is passed to regex_replace that would be useful to know. And, if it is not passed, that would be a useful addition to the regex capabilities provided by Publisher, as a future enhancement (possibly via an option in the Find and Replace regex options).

 

Edited February 6, 2023 by walt.farrell
Corrected syntax so it matches one version of what Boost allows

[laurent32]

38 minutes ago, walt.farrell said:

My research continues

@walt.farrell, I love the way you dive deep into things !

I must say = "respect" and thanks for sharing all that knowledge !

[walt.farrell]

You're welcome, @laurent32.

Further research notes:

1. The format for conditional replacement strings in Boost is specified at https://www.boost.org/doc/libs/1_81_0/libs/regex/doc/html/boost_regex/format/boost_format_syntax.html
2. I have tried that format (and edited my message above to show a correct format) in another application which uses Boost (Notepad++) and the conditional replacement works there.
3. I have again tried it in Publisher, using a correct format, and it does not work. Therefore, it is most likely that Publisher does not enable this formatting option when issuing the regex_replace call to process the replacement string. Thus ends the research, with the answer that what @Pyanepsion wants to accomplish does not seem to be possible, currently, in Publisher.

 

[Pyanepsion]

@walt.farrell, What search and replace routines do you use in Notepad++?

[walt.farrell]

5 hours ago, Pyanepsion said:

What search and replace routines do you use in Notepad++?

I didn't actually try your case, but I think it should be:

Find: (\s(?=(:|»|%)))|(«)\s

Replace: (?1<nnbsp>:\2<nnbsp>)

(But you would probably need the actual nnbsp character in the replacement string.)

[Andreas CH]

Hallo Walt

16 minutes ago, walt.farrell said:

I didn't actually try your case, but I think it should be:

Find: (\s(?=(:|»|%)))|(«)\s

Replace: (?1<nnbsp>:\2<nnbsp>)

(But you would probably need the actual nnbsp character in the replacement string.)

Bei deinem Ausdruck wird zu viel markiert.

Il lui dit : « je sais. C’est 15 %. »

Es braucht ?<= vor «

 

richtig ist:

(\s(?=(:|»|%)))|(?<=«)\s

Il lui dit : « je sais. C’est 15 %. »

 

Das Ersetzen (Replace: (?1<nnbsp>:\2<nnbsp>) ) funktioniert bei mir nicht. Ich muss immer das Zeichen im Dokument markieren, kopieren und dann im Ersetzen-Dialog einfügen.

 

 

[lacerto]

Obsolete.

[walt.farrell]

41 minutes ago, Andreas CH said:

Bei deinem Ausdruck wird zu viel markiert.

I think my specifications match what @Pyanepsion showed in his first post.

[laurent32]

48 minutes ago, Andreas CH said:

Das Ersetzen (Replace: (?1<nnbsp>:\2<nnbsp>) ) funktioniert bei mir nicht. Ich muss immer das Zeichen im Dokument markieren, kopieren und dann im Ersetzen-Dialog einfügen.

english = The replace (Replace: (?1<nnbsp>:\2<nnbsp>) ) doesn't work for me. I always have to select the character in the document, copy it and then paste it in the replace dialog

=> It doesn't work for me either.

I think that is what @walt.farrell has found :

21 hours ago, walt.farrell said:

I have again tried it in Publisher, using a correct format, and it does not work. Therefore, it is most likely that Publisher does not enable this formatting option when issuing the regex_replace call to process the replacement string. Thus ends the research, with the answer that what @Pyanepsion wants to accomplish does not seem to be possible, currently, in Publisher.

 

[walt.farrell]

4 minutes ago, laurent32 said:

english = The replace (Replace: (?1<nnbsp>:\2<nnbsp>) ) doesn't work for me. I always have to select the character in the document, copy it and then paste it in the replace dialog

I think @Pyanepsion showed what the Find/Replace fields contain when you paste in the actual character. Or when you use the pull-down for the field to insert special characters. So, click on the pull-down, and you'll find a way to insert the space. And it will look like what @Pyanepsion showed.

[laurent32]

On 2/5/2023 at 10:37 AM, Pyanepsion said:

Is this possible?

Is it possible to insert a unicode value into Regex formulas such as \u+202f, or to indicate a variable such as <nnbsp> that appears in the occurrence display?

Or, is it necessary to paste a narrow non-breaking space <nnbsp> (obtained by U+202F from the text)?

If I'm correct, question still unanswered ?

 

[laurent32]

20 minutes ago, walt.farrell said:

I think @Pyanepsion showed what the Find/Replace fields contain when you paste in the actual character. Or when you use the pull-down for the field to insert special characters. So, click on the pull-down, and you'll find a way to insert the space. And it will look like what @Pyanepsion showed.

We don't have <nnbsp> in the pull down, we only have access to <nbsp> :

And I tried the replace formula with <nbsp>, I can't make it work… ?

[lacerto]

Obsolete.

[walt.farrell]

19 minutes ago, laurent32 said:

And I tried the replace formula with <nbsp>, I can't make it work… ?

Did you try it with the character inserted from the dropdown (which would show as <nbsp>), or with a pasted non-breaking space, or with the text <nbsp>?

The pulldown or pasting would be required. But even then, if you're trying the conditional replacement, that does not work in Publisher.

[walt.farrell]

21 minutes ago, laurent32 said:

We don't have <nnbsp> in the pull down, we only have access to <nbsp> :

Thanks. Then you would need to paste in the proper character or use the \u or \x notation, I think.