fixes bug 1212088 - Expose graphics related CSV report via webapp

[peterbe]

@AdrianGaudebert r?
cc @twobraids

I'd like to get some early eyes on this whilst I wait for the Graphics team to get back to me about which columns we can blanket out. E.g. adu_count.
Once we have the most expensive columns blanket'ed out (e.g. adu_count and duplicate_of) that query will become many many times faster.

How to test this? Run the branch and load http://socorro.dev/graphics_report/?date=2015-10-08 for example. Note that you need to be logged in.

How to test it on the command line...
First you need to generate an API token that has the "Run long queries" permission.
Then something like this:

curl -v -H "Auth-Token: 12312312312312312" http://socorro.dev/graphics_report/?date=2015-10-08

Or if you want to test that it gets gzipped properly, download it like this:

curl --compressed -v -H "Auth-Token: 12312312312312312" http://socorro.dev/graphics_report/?date=2015-10-08 > /tmp/reports.csv

And then you should be able open /tmp/reports.csv in Excel or Quip or Libreoffice or whatever you prefer.

[peterbe]

@AdrianGaudebert Note in the last iteration that I put adu_count sub-select back in because they have stated that they need that.
I suspect that if we can change this query to reports_clean AND switch to using product_adi I think this query will perform a lot better.
Having said that, this URL will require a token and I've prepared the graphics team about that.

So, I suggest we land this (pending review of course!) and evaluate what to do once we see if this is hurting our PG.

[adngdb]

According to the doc, you can write ARRAY[] instead, but I haven't tested it.

[peterbe]

breakpad=> select ARRAY[] As things;
ERROR:  cannot determine type of empty array
LINE 1: select ARRAY[] As things;
               ^
HINT:  Explicitly cast to the desired type, for example ARRAY[]::integer[].
breakpad=> select ARRAY(select 1 where false) As things;
 things
--------
 {}
(1 row)

[peterbe]

I tried something like that first but it didn't work. The reason I ended up with ARRAY(SELECT 1 WHERE FALSE) was because I found some stackoverflow post that tried to answer the same question I think I had.

[adngdb]

r+