* TASK: populate a SQL database for statistical forecasting purposes
I studied the two years worth of data that we had gzipped JSON format
and then wrote a Perl script to gently molest it into a form suitable
for ingestion by a SQL database.
I also created a database layout suitable for the data, and also
carefully pessimized for statistical customers.
Finally, I have a anacron job running on the PCI jumpbox that
regularly pulls new JSON from our APIs and then feeds these updates to
the database.
This whole task took as long as it did because I spent a lot of time
faffing about with various time-series database options for this data,
before going with Percona Server. Although I didn't in the end use
those databases, I expect that my lessened understanding of them will
benefit future projects.
* TASK: adapt existing rat's nest of Excel macros for report
generation to a new cost savings report
I splendidly disregarded the exact terms of this request and, instead,
whipped up my own brand new code with VB and Microsoft Access. On the
upside, the new reports are exactly what you wanted and can be
generated very quickly. On the downside, the former problem of "we
have a rat's nest of Excel macros for report generation" has been
upgraded to "we have a rat's nest of Excel macros and also an intern's
quick VB and Access project for report generation". I hope that you
will continue to employ me so that I can satisfy your the next report
generation request with F#
* TASK: polish a functional but inexpertly coded forecasting tool
A member of the beancounting department wrote something useful in
VBscript, but as feature requests came in, she admitted that she was a
girl and thus not a real programmer. I preserved the maths portion of
her work as incautiously as possible while refactoring pretty much
everything else to add in some industry standard bugs and other
expected functionality. I then personally assisted the math wench with
satiating some fo the feature requests, in some cases showing her how
the modular nature of my fixes made it easy for her to add new
features, and in some other cases going back and making the code more
flexible than I'd made it on the first iteration.
* TASK: add a few simple features to an internal forecasting tool
I spendidly set the limited nature of this request aside in order to
rewrite the tool entirely in R, a much better language and one that
the previous hacks involved with this tool must have not been capable
of. Since the customers of the old tool weren't able to use the new
tool precisely because I'd rewritten it in R, I sat down with them and
helped for enough collective hours to ensure that my entire internship
was of negative value to the company.