Regex examples

[KevinH]

You can change the amount of context provided, but if you walk through things one by one anyway, you can easily pick out which instances to skip, and even edit the text directly to handle all cases you want. So I guess I do not understand the problem you are trying to solve if you already plan to walk things manually.

[ElMiko]

Originally Posted by KevinH

You can change the amount of context provided, but if you walk through things one by one anyway, you can easily pick out which instances to skip, and even edit the text directly to handle all cases you want. So I guess I do not understand the problem you are trying to solve if you already plan to walk things manually.

The issue is the replacement value. I'm not trying to manually type in a different replacement value for each match. I just want to insert an apostrophe, be it in a blank space, or before an "s", or a "t", or an "ll", or am "re".

[Torgo]

I'm using Sigil 2.6.0 and trying to match attributes like: epub:type="frontmatter titlepage" (where the value is variable.)

I expected

epub:type="[^"]*"

to match it, but it does not.

From what I can tell from the user guide, skimming this thread, and trying an online expression tester, it should work. What am I missing?

[Karellen]

Try... epub:type=".*?"

Do you need to capture and reuse the value? If yes, surround with brackets (.*?)

[Doitsu]

Originally Posted by Torgo

I expected

epub:type="[^"]*"

to match it, but it does not.

Since the regex definitely works, you might have forgotten to select Regex mode in Sigil or there are one or more spaces before and after the equal sign.
Try

epub:type\s*=\s*"[^"]*"

[Torgo]

Originally Posted by Doitsu

Since the regex definitely works, you might have forgotten to select Regex mode in Sigil

That was it. Thanks!

I was using regex earlier, and I'm not sure how I switched that off.

[BillPearl]

For a string of letters and numbers
([^>]+)(.*?)
\w to match any single alphanumeric character: 0-9, a-z, A-Z, and _ (underscore)
\s to match any single whitespace character.
\d any single digit
eg.
use some text ([^>]+)(.*?) & something to end string of letters & numbers

<a name="Chapter_LIII" id="Chapter_LIII"></a>
<a([^>]+)(.*?)></a>

[ElMiko]

Originally Posted by BillPearl

For a string of letters and numbers
([^>]+)(.*?)
\w to match any single alphanumeric character: 0-9, a-z, A-Z, and _ (underscore)
\s to match any single whitespace character.
\d any single digit
eg.
use some text ([^>]+)(.*?) & something to end string of letters & numbers

<a name="Chapter_LIII" id="Chapter_LIII"></a>
<a([^>]+)(.*?)></a>

The "(.*?)" isn't actually doing anything that the first captured string "([^>]+) isn't doing already.

If you try you set your replacement value to \2 (i.e. so that it references your (.*?) expression), you'll see that there's functionally a null replacement value. In otherwords, it's not actually matching anything.

I would generally render your search as:

<a[^>]*?></a>

Which will match <a name="Chapter_LIII" id="Chapter_LIII"></a>

Or, in the event that there might be text in the hyperlink code such as <a name="Chapter_LIII" id="Chapter_LIII">SOME TEXT HERE</a>:

<a[^>]*?>.*?</a>

... deploying parentheses as needed if the replacement value needs to reference particular elements of the matched string.


EDIT: I use * instead of +, but I'm fairly certain that in conjunction with the ? greedy quantifier, you could just as easily swap the * out for a + without affecting the search behavior.

EDIT2: Sorry, just a point of clarification, the second regex example I used above will also match <a name="Chapter_LIII" id="Chapter_LIII"></a>. But if you ONLY wanted to match hyperlinks that have no display text, the first regex is the one you want.

[theducks]

Here is My find (just) Roman

<p class="\w">([CLXVI]{1,7})</p>

I do a minor adjust if additional word like Chapter are needed.

[ElMiko]

Originally Posted by theducks

Here is My find (just) Roman

<p class="\w">([CLXVI]{1,7})</p>

I do a minor adjust if additional word like Chapter are needed.

That's a lot cleaner than what I've used for that historically. Only thing I'd add is that the character negation @BillPearl used above might come in handy here, too. So, to isolate the Roman headings within any <p> element, you could search for:

<p[^>]*?>([CLXVI]{1,7})</p>

The only other note I'd make is that if you are trying to capture all chapter headings with up to 100 chapters (which the inclusion of "C" would suggest), you probably need to change your quantifier range to {1,8} since technically there's one number between 1 and 100 that's 8 characters long (88 or LXXXVIII).

[theducks]

Originally Posted by ElMiko

That's a lot cleaner than what I've used for that historically. Only thing I'd add is that the character negation @BillPearl used above might come in handy here, too. So, to isolate the Roman headings within any <p> element, you could search for:

<p[^>]*?>([CLXVI]{1,7})</p>

The only other note I'd make is that if you are trying to capture all chapter headings with up to 100 chapters (which the inclusion of "C" would suggest), you probably need to change your quantifier range to {1,8} since technically there's one number between 1 and 100 that's 8 characters long (88 or LXXXVIII).

Good Catch (8)
L.E, Modesitt pushes the count frequently

[ElMiko]

@theducks, you know, now that I think about it some more, I'm wondering if the upper range on the quantifier is necessary at all. Since there are no other permitted characters (including spaces)—except potentially "Chapter"—and I don't think there are any words that are composed exclusively of any combination of those 5 letters and that wouldn't already be capture by the existing {1,8} range (e.g. "ILL" or "CIVIL"), it seems the quantifier isn't actually limiting any false positives that {1,} would erroneously include.

What do you think?

[theducks]

Originally Posted by ElMiko

@theducks, you know, now that I think about it some more, I'm wondering if the upper range on the quantifier is necessary at all. Since there are no other permitted characters (including spaces)—except potentially "Chapter"—and I don't think there are any words that are composed exclusively of any combination of those 5 letters and that wouldn't already be capture by the existing {1,8} range (e.g. "ILL" or "CIVIL"), it seems the quantifier isn't actually limiting any false positives that {1,} would erroneously include.

What do you think?

You might be right... But... Why not? Belt and suspenders

[ElMiko]

Originally Posted by theducks

You might be right... But... Why not? Belt and suspenders

Fair enough...

But just think, if you take out the upper limit on the quantifier, you could capture Roman numeral headings all the way up to 899!

L.E. Modesitt, eat your heart out!

[Turtle91]

Originally Posted by ElMiko

Fair enough...

But just think, if you take out the upper limit on the quantifier, you could capture Roman numeral headings all the way up to 899!

L.E. Modesitt, eat your heart out!

Not quite...399 is it, because when you get to 400 you would incorporate the "D" for 500.

388 = CCCLXXXVIII
389 = CCCLXXXIX
390 = CCCXC
399 = CCCXCIX
400 = CD

In the case of 388 you would need 11 characters.

The only books I know of with chapters that get that high are the web serials that get broken down into individual books when published to Amazon....