Re: suggested addition to 'date'

From: Giorgos Keramidas <keramida_at_ceid.upatras.gr>
Date: Tue, 15 Aug 2006 04:53:28 +0300
On 2006-08-14 11:05, Julian Elischer <julian_at_elischer.org> wrote:
>Dag-Erling Sm?rgrav wrote:
>>Julian Elischer <julian_at_elischer.org> writes:
>>>At various times I've wanted to add timestamps to logfiles as
>>>they are generated..
>>
>> while read line ; do echo $(date '+%Y-%m-%d %H:%M:%S') "$line" ; done
>
> while this is true, if you had read the rest of the thread you
> would have seen the words "without running date for every
> line".

That's true, but a bit harsh.  Julian, we have probably caused
you an inordinate amount of trouble and wasted too much of your
time already with this, so I understand why you would be
frustrated about replies like the one from DES.

I'm sorry if I have, by my replying to the thread, caused you
some of the frustration (especially since I initially thought
this was cool and a nice feature to have and changed my mind
later).  Please bear with this last message from me about
modifying date(1), and then I'll shut up until this issue is
resolved by people who know more than me about src/ stuff :-)

> I even said that it could be done using shell if one WAS
> willing to run date for each line.

Yes.  But that would be horribly inefficient.

After a fair bit of thought, my reluctancy to agree that a patch
for date(1) is in order is based on at least the following
reasons:

    - Making such a change would be a diversion from what date(1)
      traditionally does.  It's a new feature and it would
      probably be useful, but I have doubts about its usefulness
      as an inline part of date(1).

    - There are countless ways to do the same thing with a
      filter, some of which also have timestamp granularity far
      better than what date(1) can ever hope to provide (at least
      without adding custom strftime(3) format specifiers, which
      is a needless diversion from standards).

      For example, a Perl-based filter can use gettimeofday() and
      provide timestamps with sub-second accuracy.  This is not
      (currently) possible with strftime(3), which is what
      date(1) is based on :(

    - Now that this thread has been spawned, people who want to
      do something like this can easily find several suggestions
      to satisfy their needs, without modifying date.  Admittedly
      this is a very weak argument for *NOT* adding new features,
      but still a thought that raises a few doubts about the
      immediate usefulness of the feature.

Just my $0.02 on the topic.  If you still think it's a good thing
to have and somebody(TM) comes up with a patch that doesn't cause
people too many problems, feel free to ignore everything above :)

Regards,
Giorgos
Received on Mon Aug 14 2006 - 23:54:21 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:59 UTC