Re: psql: show only failed queries

On Wed, Jul 9, 2014 at 9:06 PM, Pavel Stehule <pavel.stehule@gmail.com> wrote:
> Hi
>
>
> 2014-07-09 7:07 GMT+02:00 Fujii Masao <masao.fujii@gmail.com>:
>
>> On Mon, Jun 30, 2014 at 8:33 PM, Pavel Stehule <pavel.stehule@gmail.com>
>> wrote:
>> >
>> >
>> >
>> > 2014-06-30 13:01 GMT+02:00 Abhijit Menon-Sen <ams@2ndquadrant.com>:
>> >
>> >> At 2014-06-30 12:48:30 +0200, pavel.stehule@gmail.com wrote:
>> >> >
>> >> > +      <para>
>> >> > +      Print a failed SQL commands to standard error output. This is
>> >> > +      equivalent to setting the variable <varname>ECHO</varname> to
>> >> > +      <literal>errors</literal>.
>> >>
>> >> No "a", just "Print failed SQL commands …".
>> >>
>> >> > -        <option>-e</option>.
>> >> > +        <option>-e</option>. If set to <literal>error</literal> then
>> >> > only
>> >> > +        failed queries are displayed.
>> >>
>> >> Should be "errors" here, not "error".
>> >>
>> >> >       printf(_("  -a, --echo-all           echo all input from
>> >> > script\n"));
>> >> > +     printf(_("  -b  --echo-errors        echo failed commands sent
>> >> > to
>> >> > server\n"));
>> >> >       printf(_("  -e, --echo-queries       echo commands sent to
>> >> > server\n"));
>> >>
>> >> Should have a comma after -b to match other options. Also I would
>> >> remove
>> >> "sent to server" from the description: "echo failed commands" is fine.
>> >
>> >
>> > fixed
>>
>> $ psql -b
>> bin/psql: invalid option -- 'b'
>> Try "psql --help" for more information.
>>
>> I got this error. ISTM you forgot to add 'b' into the third argument of
>> getopt_long in startup.c.
>>
>
> fixed
>
>>
>>          <application>psql</application> merely prints all queries as
>>          they are sent to the server. The switch for this is
>> -        <option>-e</option>.
>> +        <option>-e</option>. If set to <literal>errors</literal> then
>> only
>> +        failed queries are displayed.
>>
>> I think that where failed queries are output should be documented here.
>> Otherwise users might misunderstand they are output to standard output
>> like ECHO=all and queries do.
>>
>> It's better to add "The switch for this is <option>-b</option>." into the
>> doc.
>
>
> fixed
>
>>
>> +    else if (strcmp(prev2_wd, "\\set") == 0)
>> +    {
>> +        if (strcmp(prev_wd, "ECHO") == 0)
>> +        {
>> +            static const char *const my_list[] =
>> +            {"none", "errors", "queries", "all", NULL};
>> +
>> +            COMPLETE_WITH_LIST_CS(my_list);
>> +        }
>> +        else if (strcmp(prev_wd, "ECHO_HIDDEN") == 0)
>> +        {
>> +            static const char *const my_list[] =
>> +            {"noexec", "off", "on", NULL};
>> +
>> +            COMPLETE_WITH_LIST_CS(my_list);
>> +        }
>> +    }
>>
>> I think that adding tab-completions of psql variables is good, but
>> adding those of only ECHO and ECHO_HIDDEN seems half-baked.
>> Probably this part should be split into separate patch.
>
>
> fixed
>
> please, see updated patch in attachment

Thanks for updating the patch!

Barring any objection, I will commit this patch except tab-completion part.

I'm not against adding tab-completion support for psql variables, but I'm
not sure if it's good idea or not to treat only one or two variables special
and add tab-completions for them. There are other variables which accept
special argument (e.g., COMP_KEYWORD_CASE) and it's basically worth
adding tab-completion support for them.

Regards,

--
Fujii Masao