Regex help anyone?

[seanos]

I had sort of got used to Sigil’s regex but I’m finding myself a bit lost with Calibre’s different version.

The book I’m editing has a lot of redundant tags like...

<div class="text">CONTENT</div>

...where I’d just like to remove the div or ...

<p class="normal-262-0-override"><span class="no-style-override-9">CONTENT</span></p>

...which I’d like to be...

<p>CONTENT</p>

Can any kind soul help out?

[Perkin]

IF the div or p classes match your examples you can use the following

Search

<div class="text">([^<]+)</div>

or

<p class="normal-262-0-override"><span class="no-style-override-9">([^<]+)</span></p>

Replace

<p>\1</p>

Edit: Will only work if there are no <tags> in the CONTENT portion, if there is we can do alternate searches, by changing the ([^<]+) to (.+?)

[seanos]

Thanks Perkin, the para one works fine but “No matches were found” for the div.

Here’s an example. This one I can just delete, but the others (uselessly as the class isn’t even defined) wrap a lot of text.

			<div class="text">
				<p class="normal-262-0-override"><span class="no-style-override-9"><br/></span></p>
				<p style="page-break-after:always;"></p><div style="page-break-after:always"></div><p class="normal-262-0-override"><span class="no-style-override-9"><br/></span></p>
			</div>

[seanos]

Hmm...seems like there are some of the paras it won’t find either.

I even copied the match text from the ones I can see in the doc, but it still won’t find them.

[Thasaidon]

The Sigil forum has a sticky thread for regex examples. As the Editor's version of Regex is different, it may be helpful to have a similar sticky in the Editor forum.

[mrmikel]

One thing to watch out for is that the case sensitive check box will override the search criteria. If it is NOT checked then [a-z] will find not only a-z but A-Z as well.

Are you searching inside a table? It seems to have trouble doing that.

Also, are you typing things into the search or are you copying and pasting. Copying and pasting, then altering will catch hidden characters, like non-breaking spaces which might be present.

[seanos]

Case sensitive was not checked. Not searching in a table. I copied and pasted to test just that possibility (hidden chars) but could not find the copied text.

[seanos]

Ah...for the paras, the cases where it works are ones where there are no other tags in the paragraph. Is that a clue?

[mrmikel]

How about one step at a time? Do the inner ones first and the outer ones after the first are gone.

Also, some symbols like . are used in regex commands. If you suspect they might be, use \ in front of them, like \.

[seanos]

Some paragraphs will always have tags in them though. I can try though. Either way I’ll get there in the end; it would just be quicker (and require less hunting down stray </span>s) with regex.

[mrmikel]

The thing is if you have trouble figuring out how to separate the two, how is a program or regex search supposed to figure it out. The perfect one that will get everything you want always seems to sweep up others too.

My solution is to gear for these ones that sweep everything and search one at a time, ready to hit replace and find, but also ready to go up to the text to fix something that would be overfixed by the regex. It is not so fast, but it is much faster than reading and fixing line by line or faster than repairing the damage of an overactive replace.

[seanos]

Well, that is what I’ve been doing more or less. Most of the formatting in this book is utterly useless anyway.

I don’t actually have any trouble at all identifying the required sequences of characters; it’s just the suggested regex doesn’t seem to work consistently. The exact sequences of chars copied from the document are not found with regex though they are with a normal search. This leads me to wonder if there is something in the syntax that’s preventing it from working, but since it’s a bit of a black art to me and I don’t even know what the flavour of regex in Calibre is called, I don’t really know what to look for.

It doesn’t seem too unreasonable to want to search for [exact sequence A]SOME TEXT[first occurrence of exact sequence B].

[mrmikel]

It seems somewhat inconsistent to me too. But my abilities are kind of like yours. There are various cheatsheets that cover it, but it is not something that is easy to understand. So I don't know what I don't know.

If you run the search again, sometimes it seems to work where it did not scoop it up the first time. It seems like they are the same too.

[DiapDealer]

I've found there's really not that much difference between calibre-edit's regex engine and Sigil's. And most of those differences are very "edge case" kind of stuff: searching for unicode code-points, and the like. Mostly, calibre-edit's regex engine just adds a few things that weren't in Sigil's: like allowing variable-length lookbehinds, and matching the start/end of words with \m \M.

If case matters ... then case matters. And if so, the appropriate box needs to be checked.

[seanos]

Figured out what I wanted was this...

<p class="normal-262-0-override"><span class="no-style-override-9">(.*)</span></p>

<p>\1</p>