Quantcast

Handling Missing Data

Get the WebProNews Newsletter:


[ Business]

I have an old Perl project that goes out to a Government web site, ftp’s some files, massages them in various ways, and spits out some output. Over time, the project grew, and now does more than it used to.

To keep it generic and simple, let’s pretend that originally it went out and got “temperature” files every hour. That was simple enough: about the only error condition would be the inability to get one or more of the files it wanted. My program simply kept track of what it actually got and only ran the rest of the process on new files.

The section of code that does that isn’t hard:

foreach(@file_to_get) {
        $newerr=0;
        $ftp->get($f,$f) or $newerr=1;
        push @gotfiles, "$f" if not $newerr;
        }

            foreach(@gotfiles) {
                ...
        }

Well, then “precipitation” was added, and then each file might have processing for different “customers”. A bit of extra looping, no big deal.. but sometimes we’d get a “short file” output.

Examination of logs revealed the problem: every now and then a “temperature” file would be unavailable at the ftp site, so the final output only contained “precipitation” data. From my point of view, that’s perfectly reasonable: the data wasn’t available at the input, so it’s not in the output.

But the customer doesn’t like this. “Fix it”, they said (well, much more pleasantly, but effectively that). So the question I put back to them is “How?”. In this situation, there are really several possibilities: we could:

Leave it as it is now. I guess that doesn’t count toward “fixing it”.

Don’t publish the file if it is “short”. That ignores the data that we were able to get; but if the ultimate result is pointless without ALL of it, then this makes sense.

Fill the missing slots with data tagged as “bad”. That might be reasonable too, because sometimes data we do get is tagged that way already.

Use an older “temperature” file – these things get published every hour, so we could produce a complete file using older data – I don’t know if that makes sense to the final consumer, but we could do that. If acceptable, the next question would be “how old is still acceptable?”

Programmers have to deal with stuff like this any time there is a possibility of missing inputs. I talked previously about missing fields in a tab delimited file at Handling missing data in inputs; this is a slightly different example.

*Originally published at APLawrence.com

Reddit | Furl

Bookmark WebProNews:

A.P. Lawrence provides SCO Unix and Linux consulting services http://www.pcunix.com

Handling Missing Data
Comments Off
Top Rated White Papers and Resources

Comments are closed.

  • Join for Access to Our Exclusive Web Tools
  • Sidebar Top
  • Sidebar Middle
  • Sign Up For The Free Newsletter
  • Sidebar Bottom