Re: proposal: function parse_ident

2016-03-10 15:34 GMT+01:00 Teodor Sigaev <teodor(at)sigaev(dot)ru>:

> select
>>
>> parse_ident(E'X\rXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX');
>>
>>
>> I am sending updated patch - I used json function for correct escaping -
>> the
>> escaping behave is same.
>>
>
> Hmm, it doesn't look so:
> % select parse_ident(E'_\005');
> ERROR: identifier contains disallowed characters: "_\u0005"
> % select parse_ident(E'\005');
> ERROR: missing identifier: "\u0005"
>
> but
>
> # select parse_ident(E'"\005"');
> parse_ident
> -------------
> {\x05}
>
> Error messages above point wrong character wrongly.
>
> One more inconsistence:
> # select parse_ident(E'"\005"') as "\005";
> \005
> --------
> {\x05}
>
> Display outputs of actual identifier and parse_indent are differ.
>
> Actually, I can live with both but are any other opinions? Seems, at least
> difference of actual identifier and output of parse_indent should be
> pointed in docs.

I afraid so I cannot to fix this inconsistency (if this is inconsistency -
the binary values are same) - the parameter of function is raw string with
processed escape codes, and I have not any information about original
escape sequences. When you enter octet value, and I show it as hex value,
then there should be difference. Buy I have not information about your
input (octet or hex). I have the original string of SQL identifier inside
parser, executor, but I have not original string of function parameter
inside function (not without pretty complex and long code).

I am trying describe it in doc (I am sorry for my less level English) in
new patch. Fixed duplicated oid too.

Regards

Pavel

>
> --
> Teodor Sigaev E-mail: teodor(at)sigaev(dot)ru
> WWW:
> http://www.sigaev.ru/
>