A regex function to number a mathematical ebook

[dmonasse]

The search and replace tool with regex function is really fantastic. My little society is building mathematical ebooks from latex sources. One of my problems for converting such books is that latex auto-numbers chapters, sections, subsections and theorem-like assertions (theorems, propositions, lemmas, definitions, corollaries and so on). I would like to do such a numbering in my ebook.

A solution is the following:

1) Converting from latex, I put chapters, sections, subsections and assertions in a <div> tag with a html5 data-type attribute. For example, a latex section

\section{History of the Fermat-Wiles theorem}

is converted into

<div class="section" data-type="section">History of the Fermat-Wiles theorem</div>

and

\begin{theorem}Abracadabra\end{theorem}

is converted into

<div class="theorem" data-type="theorem">Abracadabra</div>

Nota: I can't use the class attribute to denote the type of the div because the conversion process from HTML to ePub by Calibre modifies these attributes and class="theorem" may be changed into class="pcalibre25". That's the reason for the data-type attribute.

2) After conversion from latex to html (not so easy!!!) and from html to epub (easy with Calibre), I number the whole book with the Calibre editor using the search and replace tool with regex function.
The search pattern I use is:

<div.*?data-type="(chapter|section|subsection|theorem|proposition|lemma|definition|corollary)"[^>]*>

and the regex function may be:

def replace(match, number, file_name, metadata, dictionaries, data, functions, *args, **kwargs):
    if number==1: #initialization of the counts
        data['chapter']=0
        data['section']=0
        data['subsection']=0
        data['assertion']=0
    the_type=match.group(1)
    if the_type=='chapter': # begins a chapter, reinitialize the counts
        data['section']=0
        data['subsection']=0
        data['assertion']=0
        data['chapter']+=1
        return match.group()+"<span class='chapter_num'>Chapter "+str(data['chapter'])+".</span> "
    elif the_type=='section': # begins a section, reinitialize the subsection count
        data['subsection']=0
        data['section']+=1
        return match.group()+"<span class='section_num'>Section "+str(data['section'])+".</span>" 
    elif the_type=='subsection':
        data['subsection']+=1
        return match.group()+"<span class='subsection_num'>Subsection "+str(data['section'])+"."+str(data['subsection'])+".</span>"
    else: # this is an assertion
        data['assertion']+=1
        return match.group()+"<span class='assertion_num'>Assertion "+str(data['chapter'])+"."+str(data['assertion'])+".</span>"
    return ''

replace.file_order = 'spine'

Adapt the code according to your needs or wishes, this is only an example; it would be nicer to replace "Assertion" by "Theorem", "Proposition", "Lemma", "Corollary", "Definition" (very easy to do starting from the "the_type" variable). I obtain such a numbering:

Chapter 1
     Section 1
         Subsection 1.1
             Assertion 1.1
             Assertion 1.2
         Subsection 1.2
            Assertion 1.3
     Section 2
         Subsection 2.1
             Assertion 1.4
             Assertion 1.5
         Subsection 2.2
            Assertion 1.6
Chapter 2
     Section 1
         Subsection 1.1
             Assertion 2.1
             Assertion 2.2
         Subsection 1.2
            Assertion 2.3
     Section 2
         Subsection 2.1
             Assertion 2.4
             Assertion 2.5

Hope this may help. Any improvement will be welcome (even in my bad English syntax).

[kovidgoyal]

Cool I was indeed inspired by LaTeX's auto-numbering when designing the function mode feature (I used to use LaTeX extensively when I was a physicist).

[arspr]

Great usage example. Maybe you could port and post it in the saved searches sticky thread. It could be, no, it will be, a great addition.

Nevertheless:

Originally Posted by dmonasse

Nota: I can't use the class attribute to denote the type of the div because the conversion process from HTML to ePub by Calibre modifies these attributes and class="theorem" may be changed into class="pcalibre25". That's the reason for the data-type attribute.

Are you sure? I haven't tested with an HTML to epub conversion, but in an epub to epub conversion, "calibreXX" classes only appear when the original element has no given class in the source file. I've just made a quick test and I'm seeing preserved <p class="salto1">, <blockquote class="asangre"> or my own <span class="nw">.

[dmonasse]

Originally Posted by arspr

Great usage example. Maybe you could port and post it in the saved searches sticky thread. It could be, no, it will be, a great addition.

Nevertheless:


Are you sure? I haven't tested with an HTML to epub conversion, but in an epub to epub conversion, "calibreXX" classes only appear when the original element has no given class in the source file. I've just made a quick test and I'm seeing preserved <p class="salto1">, <blockquote class="asangre"> or my own <span class="nw">.

I don't understand the rules Calibre uses to flatten the CSS. I prefer to be careful and I know that the conversion process doesn't change the (html5 standard) attributes "data-xxxx".

I made a copy of this post, as suggested, in the saved searches sticky thread

Thanks for your encouragements and many thanks to Kovid for Calibre.