bug#24646: [PATCH] Complete the name of PostgreSQL databases

[Michael Mauger <michael@mauger.com>]

On Sunday, November 6, 2016 9:47 AM, Simen Heggestøyl <simenheg@gmail.com> wrote:
>On Sun, Oct 23, 2016 at 11:50 PM, Michael Mauger <michael@mauger.com> 
>wrote:
>> In the meantime, I'll review your patch and test further; I'll get 
>> back to you if I have any issues.
>
>Thanks, Michael. Did you get a chance to test it yet? If it saves you
>time, I can install the patch once you're happy with it.
>
>
>-- Simen


Several comments:
* Rather than hard-coding "psql", please use `sql-postgres-program' which can be customized to locate the psql executable.
* The "psql -l" command should also use the "-X,--no-psqlrc" option to avoid a configuration file overriding the field separator and changing the output format
* Rather than `(substring row (match-beginning 1) (match-end 1))' use `(match-string 1 row)'
* Return the nreverse of the result list so that the completion list is in the same order as psql lists them.
* `dolist' can specify the return value rather than having a separate expression after the loop. That is, (dolist (row (process-lines ...) (nreverse res)) ...)) is equivalent to (dolist (row (process-lines ...)) ...) (nreverse res)
* I'm concerned about the change to the `completing-read' call in `sql-get-login-ext'. Rather than `nil', I'd suggest `confirm' so that if the value isn't in the list, it must be confirmed.
* If the REQUIRE-MATCH parameter should really be something other than `t', then possibbly we should add another keyword :completion-required whose value would be used in the `completing-read' call (default to `t' to preserve current functionality).
* Thank you including the test module. I need to expand automated testing significantly and I appreciate your efforts to kickstart the effort and shame me to action :).


I greatly appreciate your submission but think we need to tighten up the code a bit before we commit it to the code base.