[bug-gawk] proposed new gawk time function strptime()

[Ed Morton]

Here's a summary of a couple of posts I put on comp.lang.awk 
(https://groups.google.com/d/msg/comp.lang.awk/FXrz0VIvz4A/k-ln4T6MEdUJ) to get 
some general feedback but now I'd like to get some feedback from people 
familiar with gawk internals.

I was just writing a script in a forum to convert a timestamp to a different 
form:

   $ cat file
   27/Aug/2014:23:58

   $ awk -f tst.awk file
   Wed, Aug 27, 2014 11:58:00 PM

   $ cat tst.awk
   {
      split($0,t,/[\/:]/)
      mthNr = (match("JanFebMarAprMayJunJulAugSepOctNovDec",t[2])+2)/3
      secs  = mktime(t[3]" "mthNr" "t[1]" "t[4]" "t[5]" 0")
      print strftime("%c", secs)
   }

when someone else posted a perl script that uses a built in function named 
strptime() that takes a string as the first argument and a description of the 
contents of that string in terms of time specifiers and returns the number of 
seconds. If that function existed in awk the whole of the above code would be 
written as:

   { print strftime("%c", strptime($0,"%d/%b/%Y:%H:%M")) }

I know we want to avoid cluttering up the awk language but time conversions are 
a VERY common problem and having to write the split()+match() with arithmetic 
or populate an array that maps month names to numbers etc. is pretty painful (I 
always have to look it up) and that strptime() seems much more like it'd be 
filling a glaring hole in the gawk time functions rather than adding on to them.

I think most of us understand the general issues around introducing core 
changes to gawk. The general issues though shouldn't stop us from having a 
reasonable discussion of a specific feature request so you can then decide if 
the benefits of this specific feature outweigh the effort of providing it.

In this case we have a suggestion to provide a function that would avoid a few 
lines of fairly gritty hand written code for the very common problem of working 
with time stamps in various formats.

The proposed function is well defined, exists in other languages and so is well 
known, provides symmetry with an existing gawk function (strftime()), does not 
impact any other part of the language, I would imagine is about as easy as it 
can get to implement since it already exists in libraries, fits in like a 
missing puzzle piece to the existing set of gawk time functions, and the only 
impact to existing scripts would be if someone had their own function named 
strptime() defined.

I'm not suggesting anything that would change the awk language, or take the 
code in a different direction, or even require any effort to define. In short 
it seems like it's a lot of pros and close to zero cons for very little effort.

So wrt this specific request - what do you think?

    Ed.