Re: Feature request: multilevel indexing for TeX

[Gavin Smith]

On Thu, Mar 07, 2019 at 01:57:08PM -0700, address@hidden wrote:
> OK. No problem. The number field must match /^([ivxdlcm]+|([0-9]+)$/
> to be a number. Otherwise it's a see/see also.

I found that text in the number field would have to be surrounded in 
braces, like this:

@entry{bbb}{{see also @address@hidden

otherwise the code in \pdfgettoks would make any spaces disappear, and 
"seealso" would appear.  This appears to work okay with texindex: the 
following in seealso.cp:

@entry{aaa}{1}{aaa}
@entry{bbb}{1}{bbb}
@entry{bbb}{{see also @address@hidden

produces the following in seealso.cps:

@initial {A}
@entry {aaa}{1}
@initial {B}
@entry {bbb}{1, {see also @address@hidden

So there might not be much required in terms of changes to texindex to 
support this.

I am not sure about the regex for matching Roman numerals is a good 
idea.  There are more features for index formatting that we haven't 
considered: page numbers having styles such as bold or italic is 
something that occurs in some indices in books or articles.  Then we 
might want e.g. @b{23} or @i{56} in the number field - that wouldn't be 
a "see also".  I can imagine an obscure bug where the word "mix" appears 
in the page number field when it does not mean 1009.  I'd prefer 
something more explicit.