Re: Add support for logging the current role

Itagaki Takahiro <itagaki.takahiro@gmail.com>

From: Itagaki Takahiro <itagaki.takahiro@gmail.com>

To: Stephen Frost <sfrost@snowman.net>

Cc: PostgreSQL-development <pgsql-hackers@postgresql.org>

Date: 2011-02-01T11:48:30Z

Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

Add support for an application_name parameter, which is displayed in
- 8217cfbd9918 9.0.0 cited
Make CSV column ordering a bit more logical.
- 230e8962f3a4 8.3.0 cited
Extend the format of CSV logs to include the additional information supplied
- 3bf66d6f1c3a 8.3.0 cited
Add virtual transaction IDs to CSVLOG output, so that messages coming from
- 77c166ba6cf6 8.3.0 cited
Provide for logfiles in machine readable CSV format. In consequence, rename
- fd801f4faa8e 8.3.0 cited

> Updated patch attached.

I think we need to improve postgresql.conf.sample a bit more, especially
the long line for #log_csv_fields = '...'. 330 characters in it!
  #1. Leave the long line because it is needed.
  #2. Hide the variable from the default conf.
  #3. Use short %x mnemonic both in log_line_prefix and log_csv_fields.
      (It might require many additional mnemonics.)
Which is better, or another idea?

On Sat, Jan 29, 2011 at 13:06, Stephen Frost <sfrost@snowman.net> wrote:
>> * log_csv_fields's GUC context is PGC_POSTMASTER. Is it by design?
>
> Doing SIGHUP would require addressing how to get all of the backends to
> close the old log file and open the new one, because we don't want to
> have a given log file which has two different CSV formats in it (we
> wouldn't be able to load it into the database...).  This was
> specifically addressed in the thread leading up to this patch...

I think it depends default log filename, that contains %S (seconds)
suffix. We can remove %S from log_filename; if we use a log per-day,
those log might contain different columns even after restart. If we
cannot avoid zigged csv fields completely, SIGHUP seems reasonable for it.

>> * What objects do you want to allocate in TopMemoryContext in
>>   assign_log_csv_fields() ?
> I just moved the switch to Top to be after those are allocated.

How about changing the type of csv_log_fields from List* to fixed
array of LogCSVFields? If so, we can use an array-initializer
instead of build_default_csvlog_list() ?  The code will be simplified.
Fixed length won't be a problem because it would be rare that the
same field are specified many times.

>> * Docs need some tags for itemized elements or pre-formatted codes.
>>   They looks itemized in the sgml files, but will be flattened in
>>   complied HTML files.
>
> Not sure what you're referring to here...?  Can you elaborate?  I'm not
> great with the docs. :/

Could you try to "make html" in the doc directory?
Your new decumentation after
| These columns may be included in the CSV output:
will be unaligned plain text without some tags.

-- 
Itagaki Takahiro