Why does to_tsvector ignores HTML script tags?

Question

What process is to_tsvector doing to strip tags and is it aware of <script> tags?

SELECT to_tsvector('simple', '<textarea>banana</textarea>'); -- 'banana':1
SELECT to_tsvector('simple', '<script>banana</script>'); -- empty

Where is it documented?

Answer 1

It's just <script> and <style> tags, the default parser is searching for these two specifically and reassigns the tokens as whitespace symbols. You can see that using ts_debug():
_{demo at db<>fiddle}

select v
     , to_tsvector('simple', v)
     , ts_debug.*
from(values('<script>banana</script>')
          ,('<style>banana</style>')
          ,('<asdf>banana</asdf>'))_(v)
cross join lateral ts_debug(v)
order by 1;

v	to_tsvector	alias	description	token	dictionaries	dictionary	lexemes
<asdf>banana</asdf>	'banana':1	tag	XML tag	</asdf>	{}	null	null
<asdf>banana</asdf>	'banana':1	asciiword	Word, all ASCII	banana	{english_stem}	english_stem	{banana}
<asdf>banana</asdf>	'banana':1	tag	XML tag	<asdf>	{}	null	null
<script>banana</script>		tag	XML tag	<script>	{}	null	null
<script>banana</script>		blank	Space symbols	banana	{}	null	null
<script>banana</script>		tag	XML tag	</script>	{}	null	null
<style>banana</style>		tag	XML tag	</style>	{}	null	null
<style>banana</style>		blank	Space symbols	banana	{}	null	null
<style>banana</style>		tag	XML tag	<style>	{}	null	null

It's just these two tags, as you can see in postgres/src/backend/tsearch /wparser_def.c:563:

static void
SpecialTags(TParser *prs)
{
    switch (prs->state->lenchartoken)
    {
        case 8:                 /* </script */
            if (pg_strncasecmp(prs->token, "</script", 8) == 0)
                prs->ignore = false;
            break;
        case 7:                 /* <script || </style */
            if (pg_strncasecmp(prs->token, "</style", 7) == 0)
                prs->ignore = false;
            else if (pg_strncasecmp(prs->token, "<script", 7) == 0)
                prs->ignore = true;
            break;
        case 6:                 /* <style */
            if (pg_strncasecmp(prs->token, "<style", 6) == 0)
                prs->ignore = true;
            break;
        default:
            break;
    }
}