elections used for statistic comparisons aren't helpful

[mhl]

The numbers page is meant to show you comparisons of the numbers with previous elections of the same "type" (broadly - this will need to be better defined in fixing this problem). This is currently sometimes unhelpful, e.g.:

e.g. 2:

The best elections to compare the 2017 general election to are those in 2015 and 2010. It might be best to have to explicitly link elections that should be compared like this. (This might be linked to: #103 )

[jf1]

This issue also suffers from the non-specific descriptors problem #67

'Copeland' would be better as 'Copeland UK Parliament by-election on February 23 2017'

If there'll be a flag for election/by-election (as I think there needs to be) then the comparisons system could consider only elections.

That said, now that DC generally lists all candidates for elections it's covering, is there any benefit from these comparisons? (I've always preferred a counter for locked posts, eg x/650 for June 8).

[andylolz]

Hacky data fix:

So, this code uses for_post_role to determine if elections are related (i.e. one is a prior version of another).

So the quick hack here is to change the for_post_role of ‘parl.2017-06-08’ to ‘Member of Parliament’ (to match GE2010 and GE2015), and then change the ‘for_post_role’ of the Brightside and Ogmore by-elections to ‘UK parliament’ (to match the three other by-elections..)

The more general problem, it seems, is that short of inspecting IDs, there doesn’t seem to be a way to tell a general election from a parliamentary by-election? Is that true?

EDIT: Just spotted @jf1 already pointed this last bit out!

If there'll be a flag for election/by-election (as I think there needs to be)