Cleaning a stylesheet of unused styles

[Jellby]

Calibre does some flattening to the CSS which, I believe, means it turns every selector into a class... Of course I'm not suggesting this should be done in Sigil, but maybe the "logic" behind that can be used to check the styles?

[Ahmad Samir]

I use Linux primarily; this is what I use sometimes; it's hackish, not so elegant and it doesn't take into account css rules for HTML tags, e.g.:
h2 {font-size: xx-large;}

so only css rules for .foo in e.g. <h2 class="foo">.

$ cd OEBPS/Text
$ pcregrep -o -h 'class=".+?"' * | sort -u | perl -p -e 's/class="//' | perl -p -e 's/("\n| )/|/' | perl -p -e 's/\|$//g'

this should produce a list of the css styles used in the html/xhtml files, something like:

ach1|acl|bmh|byline|cn|cotx|crt|crt1|crt2|custom1|da|ded|ded1|di|dia|dropcap|dt|dt1|dt3|dt4

then cd to the Styles dir and:

$ cd ../Styles
$ STRING="(ach1|acl|bmh|byline|cn|cotx|crt|crt1|crt2|custom1|da|ded|ded1|di|dia|dropcap|dt|dt1|dt3|dt4)"
$ pcregrep -N ANY -M '\.'$STRING'\s*?\{(\s*.+?)+?\s*\}' *.css

this should print all the css rules that are actually used in the book.

[roger64]

@Ahmad Samir

Thanks for providing a very nice code.

Your first command, the list of used styles in the text, represents a good part of the answer. Would it be possible to add, for each of the listed styles, the occurrences of each of them?

I mean, to get a list of this kind: |Italdroite 25|Header 12| and so on...
We could focus later on the least used items.

Spoiler:

roger@lmde64 ~/Bureau/Coups/OEBPS/Text $ pcregrep -o -h 'class=".+?"' * | sort -u | perl -p -e 's/class="//' | perl -p -e 's/("\n| )/|/' | perl -p -e 's/\|$//g'
Centrage|Chanson|frameFrame|Header|Heading|Italdroite|let|let1|let2|smcpCentrage|smcpChanson|smcpDroite|smcpIncise|smcpTypeA|smcpTypeV|Standard|Subtitleroger@lmde64 ~/Bureau/Coups/OEBPS/Text $

I did not find the second one so useful: it somewhat replicates the style sheet and is a little too verbose, or, more probably, I missed something...

[Ahmad Samir]

The second command is supposed to print only the css rules that appeared in the output of the first command... that's how it usually works here.

To get a count of the occurrences of each class, you can use something like:

for i in $(pcregrep -o -h 'class=".+?"' * | sort -u); do echo "$i $(grep -r $i | wc -l)"; done | sort -t ' ' -k2 -nr

this should print a list sorted in a descending order based on the occurrences count, something like:

class="di" 54
class="body" 52
class="toccn" 50
class="dropcap" 44
class="cotx" 44
class="cn" 44
class="acl" 14
class="p3" 6
class="crt1" 5
class="ull" 4
class="crt" 4
class="pt" 3

[roger64]

This is exactly what I had been wishing for! The only problem... I cannot get it working this time. After half a minute, the terminal freezes and I get no result.

[Ahmad Samir]

Originally Posted by roger64

This is exactly what I had been wishing for! The only problem... I cannot get it working this time. After half a minute, the terminal freezes and I get no result.

Freeze or just not giving any output? i.e. if you press Ctrl+C does the prompt return?

[roger64]

Originally Posted by Ahmad Samir

Freeze or just not giving any output? i.e. if you press Ctrl+C does the prompt return?

More exactly, after ten seconds the cursor stops blinking and nothing seems to happen anymore, even if I wait 2, 3 minutes.

I can return to the prompt with Ctrl+C at any time.

[Ahmad Samir]

Originally Posted by roger64

More exactly, after ten seconds the cursor stops blinking and nothing seems to happen anymore, even if I wait 2, 3 minutes.

I can return to the prompt with Ctrl+C at any time.

Does this give any output:
pcregrep -o -h 'class=".+?"' * | sort -u

[JSWolf]

Originally Posted by Jabby

Doing an epub to epub conversion in Calibre you will clean the style sheet of unused stuff and put it in alphabetical order to boot. I have been surprised, more than a few times, that Calibre will catch some of my typos and create new labels for me that I can then clean up.

Regards - John

Very bad idea to do ePub > ePub. Not only can it change your styles so you no longer know what's what, it can introduce other things you may not want. Don't do it. Use Sigil to clean up the code. But don't convert.

[roger64]

Immediate result:

roger@lmde64 ~/Bureau/Coups/OEBPS/Text $ pcregrep -o -h 'class=".+?"' * | sort -u
class="Centrage"
class="Chanson"
class="frameFrame"
class="Header"
class="Heading"
class="Italdroite"
class="let"
class="let1"
class="let2"
class="smcpCentrage"
class="smcpChanson"
class="smcpDroite"
class="smcpIncise"
class="smcpTypeA"
class="smcpTypeV"
class="Standard"
class="Subtitle"
roger@lmde64 ~/Bureau/Coups/OEBPS/Text $

[roger64]

Here is a solution for the above post coming straight from a Linux forum.
I have been told that problem was that grep was reading on stdin (for whatever it means...)

Spoiler:

pcregrep -o -h 'class=".+?"' * >fic_tmp.$$
for i in $(sort -u fic_tmp.$$); do echo "$i $(grep -rc $i fic_tmp.$$)"; done | sort -t ' ' -k2 -nr

Tried on one EPUB of mine.

roger@lmde64 ~/Bureau/Coups/OEBPS/Text $ pcregrep -o -h 'class=".+?"' * >fic_tmp.$$
roger@lmde64 ~/Bureau/Coups/OEBPS/Text $ for i in $(sort -u fic_tmp.$$); do echo "$i $(grep -rc $i fic_tmp.$$)"; done | sort -t ' ' -k2 -nr
class="smcpIncise" 225
class="smcpCentrage" 57
class="Header" 54
class="let" 35
class="let2" 35
class="let1" 35
class="Standard" 28
class="Centrage" 24
class="smcpTypeA" 17
class="smcpTypeV" 16
class="smcpDroite" 16
class="Subtitle" 5
class="Italdroite" 5
class="Heading" 2
class="smcpChanson" 1
class="frameFrame" 1
class="Chanson" 1
roger@lmde64 ~/Bureau/Coups/OEBPS/Text $

The result provides needed information (classes only).

[Ahmad Samir]

Originally Posted by roger64

Here is a solution for the above post coming straight from a Linux forum.
I have been told that problem was that grep was reading on stdin (for whatever it means...)

Glad it's working for you now. 'grep -r' seems to work here, but of course 'grep -rc $i' is a better/more elegant solution than 'grep -r $i | wc -l'.

[mmat1]

I made a small Excel-Tool wich reads the files in Sigil's "scratchpad" and covers most of the requested features.

Do you Work with Linux only ?

Maybe somebody find's it useful ...

-------------------------------
edit 12.10.2012: Attachment removed. Sigil has now some great new features, making this one obsolete.

[roger64]

Originally Posted by mmat1

I made a small Excel-Tool wich reads the files in Sigil's "scratchpad" and covers most of the requested features.

Do you Work with Linux only ?

Maybe somebody find's it useful ...

Looks very nice. For sure, Ahmad will get it working...

On Linux, scratchpad Sigil temp files are here: /tmp/Sigil/scratchpad. Nothing too confusing about it.
I did authorize your macros on OpenOffice, but nothing happened when I clicked on Analyze.
Maybe it's coming from OpenOffice, maybe from permissions, probably from me... Anyway, I'm used to see things never working at first try.

Which string did you use to search for the tags?

[mmat1]

Originally Posted by roger64

On Linux, scratchpad Sigil temp files are here: /tmp/Sigil/scratchpad. Nothing too confusing about it.
I did authorize your macros on OpenOffice, but nothing happened when I clicked on Analyze.

afaik youll have to create a new button in OO, since buttons are converted to pictures without function.

I wonder if this thing will work under LO/OO in Linux ...

Originally Posted by roger64

Which string did you use to search for the tags?

This is not understood.
Maybe this is a correct answer: It searches for everything which follows a < and is not preceeded by a /. Must-have tags ("html", "body") are omitted in most cases. And it searches for the contents of class=".*?"