Lua filter to automatically tag keywords for TeX indexing

Lua filter to automatically tag keywords for TeX indexing

[bapt a]

[-- Attachment #1.1: Type: text/plain, Size: 2518 bytes --]

Hi all,

I've started writing a technical book using Quarto markdown, which uses 
pandoc with Lua filters under the hood to produce a website as well as the 
publisher's pdf format (via LaTeX). 
I quite like to keep the source document as plain as possible, and I'm 
wondering if I could avoid the use of [concept]{.index}, which gets turned 
into \index{concept}, and instead write a Lua filter with my custom list of 
keywords, and have pandoc automatically match them as they appear in the 
text. 
As a proof of principle I wrote the following code (see below), which 
matches specific keywords, and reformats them as small-caps. I quickly 
realised that trailing punctuation, such as "concept, ..." will fail to 
match, so I'm using gsub to strip such punctuation before matching. It 
works, but I'm a bit worried:

- what's the overhead of such a filter, in practice? From what I 
understand, every single string element in the AST will be processed by 
gsub then tested for a match. Are Lua filters walking down the AST fast 
enough that I shouldn't worry about it? (as far as I can tell on small 
examples, it seems fine)

- assuming this idea is reasonable, I might want to do a few similar 
operations, e.g. reformatting program languages (as in this example code), 
wrapping keywords in \index{}, etc., and the exact format will often depend 
on the output target (html vs TeX etc.). Is there a better construct for 
this than successive if/else statements to look for matches? (I don't know 
much Lua)

Best regards,

baptiste

Lua filter:
-----

local text = require 'text'
local pandoc = require 'pandoc'

-- keywords to look for in the document
local langs = {"Matlab", "R", "Julia", "C++"}

function Includes(tab, val)
-- strip trailing punctuation before matching
local bare = string.gsub(val,"[%.|,|;|:]", "")

for index, value in ipairs(tab) do
if value == bare then
return true
end
end

return false
end

function Replace_langname(elem)
if Includes(langs, elem.text) then
return pandoc.SmallCaps(text.lower(elem.text))
else
return elem
end
end

return {{Str = Replace_langname}}










-- 
You received this message because you are subscribed to the Google Groups "pandoc-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pandoc-discuss+unsubscribe-/JYPxA39Uh5TLH3MbocFF+G/Ez6ZCGd0@public.gmane.org
To view this discussion on the web visit https://groups.google.com/d/msgid/pandoc-discuss/7f570676-2876-4e29-a8c0-9a765617f141n%40googlegroups.com.

[-- Attachment #1.2: Type: text/html, Size: 5243 bytes --]

Re: Lua filter to automatically tag keywords for TeX indexing

[Bastien DUMONT]

Iterating over all possible values at every string can be expensive. You can speed up the process by rewriting Includes like this:

```
local bool_tables = {}

function Includes(tab, val)
  -- strip trailing punctuation before matching
  local bare = string.gsub(val,"[%.|,|;|:]", "")
  -- The first time a given value of tab is tested in Includes,
  -- a table is added in bool_tables which contains
  -- a boolean key-value entry for each element in tab.
  if not bool_tables[tab] then
    bool_tables[tab] = {}
    for _, elem in ipairs(tab) do
      bool_tables[tab][elem] = true
    end
  end
  -- So we only have to test once for the key-value entry
  -- instead of iterating over all elements in tab
  -- every time Inclues is called.
  return bool_tables[tab][val]
end
```


Le Wednesday 02 November 2022 à 06:20:37PM, bapt a a écrit :
> Hi all,
> 
> I've started writing a technical book using Quarto markdown, which uses pandoc
> with Lua filters under the hood to produce a website as well as the publisher's
> pdf format (via LaTeX).
> I quite like to keep the source document as plain as possible, and I'm
> wondering if I could avoid the use of [concept]{.index}, which gets turned into
> \index{concept}, and instead write a Lua filter with my custom list of
> keywords, and have pandoc automatically match them as they appear in the text.
> As a proof of principle I wrote the following code (see below), which matches
> specific keywords, and reformats them as small-caps. I quickly realised that
> trailing punctuation, such as "concept, ..." will fail to match, so I'm using
> gsub to strip such punctuation before matching. It works, but I'm a bit
> worried:
> 
> - what's the overhead of such a filter, in practice? From what I understand,
> every single string element in the AST will be processed by gsub then tested
> for a match. Are Lua filters walking down the AST fast enough that I shouldn't
> worry about it? (as far as I can tell on small examples, it seems fine)
> 
> - assuming this idea is reasonable, I might want to do a few similar
> operations, e.g. reformatting program languages (as in this example code),
> wrapping keywords in \index{}, etc., and the exact format will often depend on
> the output target (html vs TeX etc.). Is there a better construct for this than
> successive if/else statements to look for matches? (I don't know much Lua)
> 
> Best regards,
> 
> baptiste
> 
> Lua filter:
> -----
> 
> local text = require 'text'
> local pandoc = require 'pandoc'
> 
> -- keywords to look for in the document
> local langs = {"Matlab", "R", "Julia", "C++"}
> 
> function Includes(tab, val)
> -- strip trailing punctuation before matching
> local bare = string.gsub(val,"[%.|,|;|:]", "")
> 
> for index, value in ipairs(tab) do
> if value == bare then
> return true
> end
> end
> 
> return false
> end
> 
> function Replace_langname(elem)
> if Includes(langs, elem.text) then
> return pandoc.SmallCaps(text.lower(elem.text))
> else
> return elem
> end
> end
> 
> return {{Str = Replace_langname}}
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> --
> You received this message because you are subscribed to the Google Groups
> "pandoc-discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send an email
> to [1]pandoc-discuss+unsubscribe-/JYPxA39Uh5TLH3MbocFF+G/Ez6ZCGd0@public.gmane.org
> To view this discussion on the web visit [2]https://groups.google.com/d/msgid/
> pandoc-discuss/7f570676-2876-4e29-a8c0-9a765617f141n%40googlegroups.com.
> 
> References:
> 
> [1] mailto:pandoc-discuss+unsubscribe-/JYPxA39Uh5TLH3MbocFFw@public.gmane.org
> [2] https://groups.google.com/d/msgid/pandoc-discuss/7f570676-2876-4e29-a8c0-9a765617f141n%40googlegroups.com?utm_medium=email&utm_source=footer

-- 
You received this message because you are subscribed to the Google Groups "pandoc-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pandoc-discuss+unsubscribe-/JYPxA39Uh5TLH3MbocFF+G/Ez6ZCGd0@public.gmane.org
To view this discussion on the web visit https://groups.google.com/d/msgid/pandoc-discuss/Y2QT/FV43helVxE0%40localhost.

Re: Lua filter to automatically tag keywords for TeX indexing

[John MacFarlane]

> On Nov 2, 2022, at 6:20 PM, bapt a <auguieba-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> 
> Hi all,
> 
> I've started writing a technical book using Quarto markdown, which uses pandoc with Lua filters under the hood to produce a website as well as the publisher's pdf format (via LaTeX). 
> I quite like to keep the source document as plain as possible, and I'm wondering if I could avoid the use of [concept]{.index}, which gets turned into \index{concept}, and instead write a Lua filter with my custom list of keywords, and have pandoc automatically match them as they appear in the text. 
> As a proof of principle I wrote the following code (see below), which matches specific keywords, and reformats them as small-caps. I quickly realised that trailing punctuation, such as "concept, ..." will fail to match, so I'm using gsub to strip such punctuation before matching. It works, but I'm a bit worried:
> 
> - what's the overhead of such a filter, in practice? From what I understand, every single string element in the AST will be processed by gsub then tested for a match. Are Lua filters walking down the AST fast enough that I shouldn't worry about it? (as far as I can tell on small examples, it seems fine)

The AST walking is very fast.  See the benchmarks at the beginning of https://pandoc.org/lua-filters.html for one example.

> - assuming this idea is reasonable, I might want to do a few similar operations, e.g. reformatting program languages (as in this example code), wrapping keywords in \index{}, etc., and the exact format will often depend on the output target (html vs TeX etc.). Is there a better construct for this than successive if/else statements to look for matches? (I don't know much Lua)

In lua you can do

string.gusb(val, [“(%l*)”], function (word)
    if indexable[word] then
      .. whatever ..
    end
  end)

This will run the function on every group of letters in the matched string.
Here I’m assuming you have a lua table indexable that maps words to true, e.g.

{ cow: true, horse: true }

That will be much faster than iterating through an array as you’re doing here.

-- 
You received this message because you are subscribed to the Google Groups "pandoc-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pandoc-discuss+unsubscribe-/JYPxA39Uh5TLH3MbocFF+G/Ez6ZCGd0@public.gmane.org
To view this discussion on the web visit https://groups.google.com/d/msgid/pandoc-discuss/ECDE1635-3DD4-4E57-8D66-E546B4742622%40gmail.com.

Re: Lua filter to automatically tag keywords for TeX indexing

[bapt a]

[-- Attachment #1.1: Type: text/plain, Size: 3022 bytes --]

Thank you both for the helpful replies; I don't fully understand how these 
boolean tables are used in Lua (clearly it works, I just don't fully get 
it), but it seems to be an important concept so I'll read up on it.

Thanks,

baptiste

On Friday, 4 November 2022 at 08:39:29 UTC+13 fiddlosopher wrote:

>
>
> > On Nov 2, 2022, at 6:20 PM, bapt a <augu...-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> > 
> > Hi all,
> > 
> > I've started writing a technical book using Quarto markdown, which uses 
> pandoc with Lua filters under the hood to produce a website as well as the 
> publisher's pdf format (via LaTeX). 
> > I quite like to keep the source document as plain as possible, and I'm 
> wondering if I could avoid the use of [concept]{.index}, which gets turned 
> into \index{concept}, and instead write a Lua filter with my custom list of 
> keywords, and have pandoc automatically match them as they appear in the 
> text. 
> > As a proof of principle I wrote the following code (see below), which 
> matches specific keywords, and reformats them as small-caps. I quickly 
> realised that trailing punctuation, such as "concept, ..." will fail to 
> match, so I'm using gsub to strip such punctuation before matching. It 
> works, but I'm a bit worried:
> > 
> > - what's the overhead of such a filter, in practice? From what I 
> understand, every single string element in the AST will be processed by 
> gsub then tested for a match. Are Lua filters walking down the AST fast 
> enough that I shouldn't worry about it? (as far as I can tell on small 
> examples, it seems fine)
>
> The AST walking is very fast. See the benchmarks at the beginning of 
> https://pandoc.org/lua-filters.html for one example.
>
> > - assuming this idea is reasonable, I might want to do a few similar 
> operations, e.g. reformatting program languages (as in this example code), 
> wrapping keywords in \index{}, etc., and the exact format will often depend 
> on the output target (html vs TeX etc.). Is there a better construct for 
> this than successive if/else statements to look for matches? (I don't know 
> much Lua)
>
> In lua you can do
>
> string.gusb(val, [“(%l*)”], function (word)
> if indexable[word] then
> .. whatever ..
> end
> end)
>
> This will run the function on every group of letters in the matched string.
> Here I’m assuming you have a lua table indexable that maps words to true, 
> e.g.
>
> { cow: true, horse: true }
>
> That will be much faster than iterating through an array as you’re doing 
> here.
>
>

-- 
You received this message because you are subscribed to the Google Groups "pandoc-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pandoc-discuss+unsubscribe-/JYPxA39Uh5TLH3MbocFF+G/Ez6ZCGd0@public.gmane.org
To view this discussion on the web visit https://groups.google.com/d/msgid/pandoc-discuss/23ccfb2a-c458-49e1-a275-dad452f5d2e3n%40googlegroups.com.

[-- Attachment #1.2: Type: text/html, Size: 3953 bytes --]