Clearing trash while converting.. finding with regular expressions

[Corbett]

I spent a few hours trying to find out how to get rid of my scanned pageheaders (without doing any manual work) in a LIT-file when converting to ePub using my Calibre.

the guides i have read focus on regular expressions and how to use them, but that wasnt my big problem...

First i looked in the preview, and saw a pattern. All header-rows had either a pagenumber in the beginning or the end, and (almost) always had blank lines before and after...

All my rows begins with <p> and end with </p>

Step 1, get the rows containing a number in the end.
<p.+\d</p>
Step 2, get the rows that begin with a number:
<p[^>]*>\d.+</p>
Ok, im lazy, so i want to combine them to weed out false positives (when i got numbers in the usual text from the book)
Step 3 combine the above with |
(<p[^>]*>\d.+</p>)|(<p.+\d</p>)

Step 4 Now to find empty rows
<p[^>]*> </p>
Step 5. And i only want those that have a "empty" row before and after.
<p[^>]*> </p>\s+((<p[^>]*>\d.+</p>)|(<p.+\d</p>))\s+<p[^>]*> </p>

Step 6. So i got rid of my false positives, but finally i want to get rid of the linebreak before and after, to get the textflow a bit nicer

</p>\s+<p[^>]*> </p>\s+((<p[^>]*>\d.+</p>)|(<p.+\d</p>))\s+<p[^>]*> </p>\s+<p[^>]*>

(Step 7 - FAILED)
So i want to use that expression and replace with a single space-character.... Unfortunately i failed there...

I didnt get my book converted with the wasted lines deleted for some reason, becase for some reason the search-replace wasnt activated...

Might it be so that Search-Replace is done after conversion, so i have to guess how the converted document looks when i search and replace in calibres conversion module?

But may be useful for those that are able to really replace things anyway :/

[Corbett]

I spent a few hours trying to find out how to get rid of my scanned pageheaders (without doing any manual work) in a LIT-file when converting to ePub using my Calibre.

the guides i have read focus on regular expressions and how to use them, but that wasnt my big problem...

First i looked in the preview, and saw a pattern. All header-rows had either a pagenumber in the beginning or the end, and (almost) always had blank lines before and after...

All my rows begins with <p> and end with </p>

Step 1, get the rows containing a number in the end.
<p.+\d</p>
Step 2, get the rows that begin with a number:
<p[^>]*>\d.+</p>
Ok, im lazy, so i want to combine them to weed out false positives (when i got numbers in the usual text from the book)
(<p[^>]*>\d.+</p>)|(<p.+\d</p>)

Now to find empty rows
<p[^>]*> </p>
And i only want those that have a "empty" row before and after.
<p[^>]*> </p>\s+((<p[^>]*>\d.+</p>)|(<p.+\d</p>))\s+<p[^>]*> </p>

So i got rid of my false positives, but finally i want to get rid of the linebreak before and after, to get the textflow a bit nicer

</p>\s+<p[^>]*> </p>\s+((<p[^>]*>\d.+</p>)|(<p.+\d</p>))\s+<p[^>]*> </p>\s+<p[^>]*>

So i want to use that expression and replace with a single space-character.... Unfortunately i failed there...

I didnt get my book converted with the wasted lines deleted for some reason, becase for some reason the search-replace wasnt activated...

Might it be so that Search-Replace is done after conversion, so i have to guess how the converted document looks when i search and replace in calibres conversion module?

But may be useful for those that are able to really replace things anyway :/

[Serpentine]

It's generally easier to do this stuff when you're outside of Calibre, do the conversion to EPUB/HTMLZ and try get things to be reasonable, once that's done then take the markup and process it by hand in something easier - RegexBuddy works well, tho I'm sure there are free alternatives, I never found one with a good spread of features. Sigil is nice for EPUB editing, however the current release has some problems with regex (tho it's fixed for the next release! - no real preview however).

It's also pretty tricky to help without a sample to work from, if you provide that, I'm sure I can work out a better way (there's a few problems with the regex there that will most likely miss things).

[theducks]

Moderator Notice
please do not fragment threads on the same topic. Use new reply or quote to the original thread