Indexing question

[KevinH]

Wow,
That is news to me! I am not a big user of index generation and so never learned that.
Thank you!

[roger64]

@Doitsu
Very nice. Thanks for your explanations.

Thanks all for your replies. My apologies for these dumb questions but, as a Linux user, I have been out of the ball game quite a long time. Now because of the kindness of eschwartz who provides a very useful and up-to-date Arch package, I can taste the goodness of the new Sigil.

[eschwartz]

Oh, you're an Arch user?

FWIW, the Arch Linux devs are very good at keeping it up-to-date, but 'tis true if you want to take advantage of bugfixes as they come in, my AUR package which builds from git is pretty handy.

[roger64]

Originally Posted by eschwartz

Oh, you're an Arch user?

Just a nephew (Antergos) but I use Arch and, from time to time Aur repos and packages.

A new question: my paperbook has one entry like this:

noir(e)

which means the words noir and noire have both to be indexed. No hierarchical meaning here.

How should I write this entry under the "Text to include" tab?

[Doitsu]

Originally Posted by roger64

How should I write this entry under the "Text to include" tab?

You can use regular expressions in the Text to include box. I.e. you could use:

Text to include: noir[es]*
Index Entries: noir

EDIT: I forgot the final asterisk; without it the regex won't mach "noir." I added it to this post.

[roger64]

I am confused. In the online help, it is written [Ww]illow which means either Willow or willow
If I use:

Text to include: noir[es]

does not mean: index only noire and noirs ?
or it will index noir noire noirs?

In fact I am looking for noir and noire. the second one is feminine.

[Divingduck]

Originally Posted by KevinH

Wow,
That is news to me! I am not a big user of index generation and so never learned that.
Thank you!

The index generator is one of the things why I use Sigil from time to time. It is the only program that take care of it outside a text editor (as fare as I know)

[Doitsu]

Originally Posted by roger64

does not mean: index only noire and noirs [..] or it will index noir noire noirs?

I made a stupid beginner's mistake in my initial post. I forgot to add the final asterisk, which excluded "noir" from the matches.

noir[es]* will match noir, noire, noirs and noires

Originally Posted by roger64

In fact I am looking for noir and noire.

if you don't want to include the masculine and feminine plural forms, simply use noir[e]*.

[roger64]

@Doitsu

Excellent! Thanks for your explanation and patience

[Jellby]

Originally Posted by Doitsu

noir[es]* will match noir, noire, noirs and noires

and noirse, and noireee, and noirsssessessese

if you don't want to include the masculine and feminine plural forms, simply use noir[e]*.

it wil also match noiree, noireee, noireeee...

I'd use noire?s? or noire? (if I'm right with the regex dialect). The ? means that the previous item (previous letter in this case) may appear only 0 or 1 time.

Or, if you don't want to think too much: noir|noire

[Doitsu]

Originally Posted by Jellby

I'd use noire?s? or noire? (if I'm right with the regex dialect).

Sigil uses PCRE. I.e. both question marks and asterisks will work for zero or more occurrences.

Originally Posted by Jellby

Or, if you don't want to think too much: noir|noire

IMHO, that would be the safest option.

@roger64: You might want to adopt Jellby's suggestion and simply search for all variants using the following syntax:

noir|noire|noirs|noires

Of course you could also simply generate one entry for each variation. For example:

Text to include Index entries
noir            noir
noire           noir
noirs           noir
noires          noir

[Jellby]

Originally Posted by Doitsu

Sigil uses PCRE. I.e. both question marks and asterisks will work for zero or more occurrences.

According to http://www.pcre.org/current/doc/html...ntax.html#SEC9:
? zero or one
* zero or more

[Doitsu]

Originally Posted by Jellby

According to http://www.pcre.org/current/doc/html...ntax.html#SEC9:
? zero or one
* zero or more

I stand corrected.

[roger64]

Originally Posted by Jellby

Or, if you don't want to think too much: noir|noire

Excellent.

I am preparing -slowly- an index file with about 500 entries. Your remarks are quite useful to help me get through.

Thank you both.

[roger64]

A practical test

Hi

I did a test using 30 entries (out of a 550 medium-size index) from a French book. It was mostly successful though some quirks seem to need to be ironed out. Please find attached here in the zip file:
- petiteliste.txt which is the text file I used to import and generate the index
- index.xhtml which is the index output that Sigil 9.2. (Linux Arch build) produced.
You can reproduce it by copying the text list in any EPUB text file. I also provided a screenshot of the index file in Sigil.

I - Two points that probably need to be considered.

1. Entries sorted out of the beginning of the alphabetical order.

For no reason than I can explain, some entries were classed right at the beginning of the alphabetical order under no heading. Here you can see it happen for Armani, but not Acton and for apparat and ascot but for not other words beginning with an a.

In the medium-sized index, this phenomenon did happen too and the entries concerned (between 5 to 10% of the total amount) seemed to be chosen randomly.

2. Letters with diacritics classed at the end of the alphabetical order.

Some words beginning with an accented letter (like âge or écossais) are sorted out under their own heading at the end of the alphabetical order. For the French language at least, they should be classed like if there was no diacritic: for example to be on the light side, in a French dictionary you can find âge between say affreux and agonie.

II - Practical tips.

3. Using exclusively the vertical bar (pipe).

In previous tries, I had problems with "Text to include" entries containing these very valid regex: (?i), [es] which were not dealt with but reproduced as such in the entry list. I dropped them totally and used exclusively the pipe (|) which works quite reliably for processing the index and can cope with all the cases I need.

I do not know if this problem comes from my Linux build or can be reproduced with other instances of Sigil.

4. Using a fixed value for the tab.

I used a Linux text editor named gedit to prepare the text file. I found handy to specify it in the Preferences to keep a fixed value for the tab (I chose 24). This allows for a more unified presentation.