Re^3: Regex: Identifying comments

[remiah]

Thanks for reply, pvaldes.

I noticed I misunderstood how to escape single quotes in sql. I thought of the case like this

insert into foo (bar,boo) values('''a', 'b');

And this is nothing surprise.

select '--foo';

I changed test data with real SQL with your advice. SQLite and Postgres accept these commands.

drop table foo;
create table foo(   --tabel foo
    bar text,       --fld name
    boo text        --again
);
--populate
insert into foo (bar,boo) values('''a', 'b'); -- insert quote's sake
insert into foo (bar,boo) values('c', 'd');   -- another line
select 'test' || 'abc' as a ;                 -- maybe you use dual wi
+th oracle?
select 'test''s' || 'abc' as a ;              -- maybe you use dual wi
+th oracle?
select '--foo';

So far, so good.

[pvaldes]

create table foo(   --tabel foo
    bar text,       --fld name
    boo text        --again
);

mmh... this is an interesting example, Yes. Perfectly legitim third type of comments, and difficult to debug with a regex... maybe you can profit that the fact that the type of data are limited with something like:

m/(\(|text|int|integer|char \(\d+\)),*\s*--(.*)$/

select '--foo';

This is not a very probable situation, but certainly is possible too. In any case this false comment is not after a semicolon, nor at the beginning of the line or inside a table, so if there are a ^\s*select in the same line you probably could safely ignore it. But then you could have something like this:

select field from mytable where field = 'text, --foo important information here about to be lost';

The safest actitude (although maybe a little paranoic) should be to isolate and examinate personally any case so special, the idea is: "if you found two - after a ' or a " and before a ";" in a line having the string "select" I want to see it personally"

You can improve your regex if you check previously for troublesome data:

select * from mitable where field1 like '%--%' or field2 like '%--%' or field3 like '%--%'... etc ;

[remiah]

umm..
No.

You will not see my regex fault with your examples. My regex will stumble with this sql.

update set bar = bar - 1 ; -- subtraction symbol may disappear.

I expected to see SQL parser solution in this thread, like this

my $p = SQLParser->new(type=>'mysql', sql=>$sql)
  or die SQLParser->error();

$p->prettyprint(1);
$p->without_comment(1);
print $p->sql;

At first I looked SQL::Parser. It seems quite near for such tasks, but I couldn't find good solution to rip off comments. Do you know such module?

[pvaldes]

use strict; use warnings;

while (<DATA>) {
    chomp; 
    next if /^\s*--/;
    print $_,"\n" if !/--/;
    elsif (/--/ && !/--(.*?);/){
         s/--(.*?)$//;
        print $_,"\n"}
    elsif (/--(.*?);\s*--(.*?)$/){
         s/\s*--$2//;
         print $_,"\n"}
    else { print $_,"\n"}
}

__DATA__
select 'text' from foo;       -- comment
select '--Not comment' from foo;        --But this is
select q from z; -- as is this
select '--Not this' + '--either' from foo;
select 'qaws' + make from "a"; -- comment with 'a' quote
select 'a' from 'b' with 'c' -- comment with 'a --' comment
-- test comment (add1)
select 'text\'s' from foo    --escaped  ... (add2)
select 'text\'s' from foo        --escaped'  ... (add3)
update set bar = bar - 1 ; -- subtraction symbol preserved
create table ( -- fo
field text, -- fufufu
field int)