PostgreSQL replace HTML entities function

Question

I've found this very interesting function on internet :

CREATE OR REPLACE FUNCTION strip_tags(TEXT) RETURNS TEXT AS $$
    SELECT regexp_replace(regexp_replace($1, E'(?x)<[^>]*?(\s alt \s* = \s* ([\'"]) ([^>]*?) \2) [^>]*? >', E'\3'), E'(?x)(< [^>]*? >)', '', 'g')
$$ LANGUAGE SQL;

But it doesn't remove html codes like: "   Is it possible to remove them using regexp_replace?

Answer 1

Yes it is possible to replace HTML or other character entities with the respective characters using a function. First create a character entity table:

create table character_entity(
    name text primary key,
    ch char(1) unique
);
insert into character_entity (ch, name) values
    (E'\u00C6','AElig'),(E'\u00C1','Aacute'),(E'\u00C2','Acirc'),(E'\u00C0','Agrave'),(E'\u0391','Alpha'),(E'\u00C5','Aring'),(E'\u00C3','Atilde'),(E'\u00C4','Auml'),(E'\u0392','Beta'),(E'\u00C7','Ccedil'),
    (E'\u03A7','Chi'),(E'\u2021','Dagger'),(E'\u0394','Delta'),(E'\u00D0','ETH'),(E'\u00C9','Eacute'),(E'\u00CA','Ecirc'),(E'\u00C8','Egrave'),(E'\u0395','Epsilon'),(E'\u0397','Eta'),(E'\u00CB','Euml'),
    (E'\u0393','Gamma'),(E'\u00CD','Iacute'),(E'\u00CE','Icirc'),(E'\u00CC','Igrave'),(E'\u0399','Iota'),(E'\u00CF','Iuml'),(E'\u039A','Kappa'),(E'\u039B','Lambda'),(E'\u039C','Mu'),(E'\u00D1','Ntilde'),
    (E'\u039D','Nu'),(E'\u0152','OElig'),(E'\u00D3','Oacute'),(E'\u00D4','Ocirc'),(E'\u00D2','Ograve'),(E'\u03A9','Omega'),(E'\u039F','Omicron'),(E'\u00D8','Oslash'),(E'\u00D5','Otilde'),(E'\u00D6','Ouml'),
    (E'\u03A6','Phi'),(E'\u03A0','Pi'),(E'\u2033','Prime'),(E'\u03A8','Psi'),(E'\u03A1','Rho'),(E'\u0160','Scaron'),(E'\u03A3','Sigma'),(E'\u00DE','THORN'),(E'\u03A4','Tau'),(E'\u0398','Theta'),
    (E'\u00DA','Uacute'),(E'\u00DB','Ucirc'),(E'\u00D9','Ugrave'),(E'\u03A5','Upsilon'),(E'\u00DC','Uuml'),(E'\u039E','Xi'),(E'\u00DD','Yacute'),(E'\u0178','Yuml'),(E'\u0396','Zeta'),(E'\u00E1','aacute'),
    (E'\u00E2','acirc'),(E'\u00B4','acute'),(E'\u00E6','aelig'),(E'\u00E0','agrave'),(E'\u2135','alefsym'),(E'\u03B1','alpha'),(E'\u0026','amp'),(E'\u2227','and'),(E'\u2220','ang'),(E'\u00E5','aring'),
    (E'\u2248','asymp'),(E'\u00E3','atilde'),(E'\u00E4','auml'),(E'\u201E','bdquo'),(E'\u03B2','beta'),(E'\u00A6','brvbar'),(E'\u2022','bull'),(E'\u2229','cap'),(E'\u00E7','ccedil'),(E'\u00B8','cedil'),
    (E'\u00A2','cent'),(E'\u03C7','chi'),(E'\u02C6','circ'),(E'\u2663','clubs'),(E'\u2245','cong'),(E'\u00A9','copy'),(E'\u21B5','crarr'),(E'\u222A','cup'),(E'\u00A4','curren'),(E'\u21D3','dArr'),
    (E'\u2020','dagger'),(E'\u2193','darr'),(E'\u00B0','deg'),(E'\u03B4','delta'),(E'\u2666','diams'),(E'\u00F7','divide'),(E'\u00E9','eacute'),(E'\u00EA','ecirc'),(E'\u00E8','egrave'),(E'\u2205','empty'),
    (E'\u2003','emsp'),(E'\u2002','ensp'),(E'\u03B5','epsilon'),(E'\u2261','equiv'),(E'\u03B7','eta'),(E'\u00F0','eth'),(E'\u00EB','euml'),(E'\u20AC','euro'),(E'\u2203','exist'),(E'\u0192','fnof'),
    (E'\u2200','forall'),(E'\u00BD','frac12'),(E'\u00BC','frac14'),(E'\u00BE','frac34'),(E'\u2044','frasl'),(E'\u03B3','gamma'),(E'\u2265','ge'),(E'\u003E','gt'),(E'\u21D4','hArr'),(E'\u2194','harr'),
    (E'\u2665','hearts'),(E'\u2026','hellip'),(E'\u00ED','iacute'),(E'\u00EE','icirc'),(E'\u00A1','iexcl'),(E'\u00EC','igrave'),(E'\u2111','image'),(E'\u221E','infin'),(E'\u222B','int'),(E'\u03B9','iota'),
    (E'\u00BF','iquest'),(E'\u2208','isin'),(E'\u00EF','iuml'),(E'\u03BA','kappa'),(E'\u21D0','lArr'),(E'\u03BB','lambda'),(E'\u2329','lang'),(E'\u00AB','laquo'),(E'\u2190','larr'),(E'\u2308','lceil'),
    (E'\u201C','ldquo'),(E'\u2264','le'),(E'\u230A','lfloor'),(E'\u2217','lowast'),(E'\u25CA','loz'),(E'\u200E','lrm'),(E'\u2039','lsaquo'),(E'\u2018','lsquo'),(E'\u003C','lt'),(E'\u00AF','macr'),
    (E'\u2014','mdash'),(E'\u00B5','micro'),(E'\u00B7','middot'),(E'\u2212','minus'),(E'\u03BC','mu'),(E'\u2207','nabla'),(E'\u00A0','nbsp'),(E'\u2013','ndash'),(E'\u2260','ne'),(E'\u220B','ni'),
    (E'\u00AC','not'),(E'\u2209','notin'),(E'\u2284','nsub'),(E'\u00F1','ntilde'),(E'\u03BD','nu'),(E'\u00F3','oacute'),(E'\u00F4','ocirc'),(E'\u0153','oelig'),(E'\u00F2','ograve'),(E'\u203E','oline'),
    (E'\u03C9','omega'),(E'\u03BF','omicron'),(E'\u2295','oplus'),(E'\u2228','or'),(E'\u00AA','ordf'),(E'\u00BA','ordm'),(E'\u00F8','oslash'),(E'\u00F5','otilde'),(E'\u2297','otimes'),(E'\u00F6','ouml'),
    (E'\u00B6','para'),(E'\u2202','part'),(E'\u2030','permil'),(E'\u22A5','perp'),(E'\u03C6','phi'),(E'\u03C0','pi'),(E'\u03D6','piv'),(E'\u00B1','plusmn'),(E'\u00A3','pound'),(E'\u2032','prime'),
    (E'\u220F','prod'),(E'\u221D','prop'),(E'\u03C8','psi'),(E'\u0022','quot'),(E'\u21D2','rArr'),(E'\u221A','radic'),(E'\u232A','rang'),(E'\u00BB','raquo'),(E'\u2192','rarr'),(E'\u2309','rceil'),
    (E'\u201D','rdquo'),(E'\u211C','real'),(E'\u00AE','reg'),(E'\u230B','rfloor'),(E'\u03C1','rho'),(E'\u200F','rlm'),(E'\u203A','rsaquo'),(E'\u2019','rsquo'),(E'\u201A','sbquo'),(E'\u0161','scaron'),
    (E'\u22C5','sdot'),(E'\u00A7','sect'),(E'\u00AD','shy'),(E'\u03C3','sigma'),(E'\u03C2','sigmaf'),(E'\u223C','sim'),(E'\u2660','spades'),(E'\u2282','sub'),(E'\u2286','sube'),(E'\u2211','sum'),
    (E'\u2283','sup'),(E'\u00B9','sup1'),(E'\u00B2','sup2'),(E'\u00B3','sup3'),(E'\u2287','supe'),(E'\u00DF','szlig'),(E'\u03C4','tau'),(E'\u2234','there4'),(E'\u03B8','theta'),(E'\u03D1','thetasym'),
    (E'\u2009','thinsp'),(E'\u00FE','thorn'),(E'\u02DC','tilde'),(E'\u00D7','times'),(E'\u2122','trade'),(E'\u21D1','uArr'),(E'\u00FA','uacute'),(E'\u2191','uarr'),(E'\u00FB','ucirc'),(E'\u00F9','ugrave'),
    (E'\u00A8','uml'),(E'\u03D2','upsih'),(E'\u03C5','upsilon'),(E'\u00FC','uuml'),(E'\u2118','weierp'),(E'\u03BE','xi'),(E'\u00FD','yacute'),(E'\u00A5','yen'),(E'\u00FF','yuml'),(E'\u03B6','zeta'),
    (E'\u200D','zwj'),(E'\u200C','zwnj')
;

This is the function:

create or replace function entity2char(t text)
returns text as $body$
declare
    r record;
begin
    for r in
        select distinct ce.ch, ce.name
        from
            character_entity ce
            inner join (
                select name[1] "name"
                from regexp_matches(t, '&([A-Za-z]+?);', 'g') r(name)
            ) s on ce.name = s.name
    loop
        t := replace(t, '&' || r.name || ';', r.ch);
    end loop;

    for r in
        select distinct
            hex[1] hex,
            ('x' || repeat('0', 8 - length(hex[1])) || hex[1])::bit(32)::int codepoint
        from regexp_matches(t, '&#x([0-9a-f]{1,8}?);', 'gi') s(hex)
    loop
        t := regexp_replace(t, '&#x' || r.hex || ';', chr(r.codepoint), 'gi');
    end loop;

    for r in
        select distinct
            chr(codepoint[1]::int) ch,
            codepoint[1] codepoint
        from regexp_matches(t, '&#([0-9]{1,10}?);', 'g') s(codepoint)
    loop
        t := replace(t, '&#' || r.codepoint || ';', r.ch);
    end loop;

    return t;
end;
$body$
language plpgsql immutable;

Use it like this:

select entity2char('HH■XXXÆYYY×ZZZ■UUU');
    entity2char     
--------------------
 HH■XXXÆYYY×ZZZ■UUU

It only works for UTF-8.

Answer 2

This classic quote may apply here: Some people, when confronted with a problem, think “I know, I'll use regular expressions.” Now they have two problems. Regex are useful, but HTML parsing is not a job they're well suited for. Jeff Atwood explains this well . To strip tags from HTML correctly some kind of parsing is necessary.

What I would recommend is that you use a more powerful PL like PL/Perl or PL/Pythonu to invoke mature and well tested HTML-stripping libraries. For example, you could use Perl's HTML::Strip via a plperl function that accepts text and returns text .

The quick and dirty way to handle this would be to use another layer of regexp_replace expressions to convert entities. This will rapidly lead you down the path alluded to by Igor though, and is best avoided by using tools that aready exist. For example, if you use HTML::Strip it'll use HTML::Entities to convert entities for you as part of the process.

Answer 3

I've been using this successfully for while - thanks for the solution. However I've just discovered that this doesn't seem to work with HTML items such as ² (superscript 2 = &sup2 ), and I suspect any other HTML item that has digit just before the closing ";".

I believe the line

from regexp_matches(t, '&([A-Za-z]+?);', 'g') r(name)

should be

from regexp_matches(t, '&([A-Za-z]+[0-9]?);', 'g') r(name)

I've tried this with a few examples and it seems to work.