Having created a new candidate with an error in the gender field (copy and paste/not paying attention problem), it's now not available to correct in the edit screen.

From @symroe

after updating her she has gone in the "known
not to be standing again" category. This is inferred by me making an edit
and not checking the 'standing in 2015' box, but I never actually stated
that she will not be standing again (I'm sure she will be, but I can't find
a source for that). Maybe we should only mark people as not standing if
there is an explicit statement to that effect?

So you can change a Wikipedia URL or Twitter username for a person, or add one where there was no previous value, but clearing the field and submitting it has no effect. Presumably we're not sending that field to PopIt in that case, and then PopIt's unexpected PUT behaviour (patching, not replacing) hits us...

In other words, only the numeric ID matters, but it makes the URLs more easily recognisable.

At the moment I'm thinking that this is potentially so open to abuse (and was problematic for YourNextMP in 2010) that we'd want to restrict the users who can do it, or have a submitted photos moderation queue.

Add a scraper or scrapers for Green Party candidates, similar to these ones:

• https://github.com/mhl/UK-Political-Parties/tree/ppc-scrapers/ppcs

Should users be able to alter how their name appears on the site? Is it possible for 2 people to register with the same user name? And does it matter?

(And do I need [an admin] to change mine?)

This should probably be a nice error message instead.

At the same time, make sure that if a person pastes in a Twitter URL when entering data, we just extract the username.

There are corrections that people might want to make to the site that they either can't figure out how to do, or aren't currently possible in the interface - the main one of these that strikes me is indicating that two different people should be merged (#25). If there was a feedback form on each page (as with Pombola) with a captcha, then we could collect these corrections to apply.

... after talking to Edmund about changing the DNS.

In the 2010 election, Graham Lee http://yournextmp.mysociety.org/person/804 unsuccessfully stood as a candidate for Ruislip, Northwood and Pinner.

The Green Party PPC page for London (http://london.greenparty.org.uk/elections/2015-general-election.html - the last (incomplete) entry) currently has him listed for neighbouring Uxbridge and South Ruislip instead.

Assuming that the Green Party page is correct, how should we handle that? Maybe a search for candidates of the same name when creating a new candidate might be part of it (there are currently 2 "Graham Lee"s in the live data as I'd forgotten he stood elsewhere last time around - sorry - http://yournextmp.mysociety.org/person/4547 is the duplicate)

On http://mark.candidates.dev.mysociety.org/person/stephen-williams/update I added a homepage url and twitter username (which appear to have saved correctly) but got the following error.

Environment:


Request Method: POST
Request URL: http://mark.candidates.dev.mysociety.org/person/stephen-williams/update

Django Version: 1.6.5
Python Version: 2.7.3
Installed Applications:
('django.contrib.admin',
 'django.contrib.auth',
 'django.contrib.contenttypes',
 'django.contrib.sessions',
 'django.contrib.messages',
 'django.contrib.staticfiles',
 'django_nose',
 'pipeline',
 'candidates',
 'debug_toolbar')
Installed Middleware:
('django.contrib.sessions.middleware.SessionMiddleware',
 'django.middleware.common.CommonMiddleware',
 'django.middleware.csrf.CsrfViewMiddleware',
 'django.contrib.auth.middleware.AuthenticationMiddleware',
 'django.contrib.messages.middleware.MessageMiddleware',
 'django.middleware.clickjacking.XFrameOptionsMiddleware')


Traceback:
File "/data/vhost/mark.candidates.dev.mysociety.org/venv/lib/python2.7/site-packages/django/core/handlers/base.py" in get_response
  112.                     response = wrapped_callback(request, *callback_args, **callback_kwargs)
File "/data/vhost/mark.candidates.dev.mysociety.org/venv/lib/python2.7/site-packages/django/views/generic/base.py" in view
  69.             return self.dispatch(request, *args, **kwargs)
File "/data/vhost/mark.candidates.dev.mysociety.org/venv/lib/python2.7/site-packages/django/views/generic/base.py" in dispatch
  87.         return handler(request, *args, **kwargs)
File "/data/vhost/mark.candidates.dev.mysociety.org/venv/lib/python2.7/site-packages/django/views/generic/edit.py" in post
  171.             return self.form_valid(form)
File "/data/vhost/mark.candidates.dev.mysociety.org/yournextmp-popit/candidates/views.py" in form_valid
  540.             person.id
File "/data/vhost/mark.candidates.dev.mysociety.org/yournextmp-popit/candidates/views.py" in set_party_membership
  254.         person = PopItPerson.create_from_popit(self.api, person_id)
File "/data/vhost/mark.candidates.dev.mysociety.org/yournextmp-popit/candidates/models.py" in create_from_popit
  79.         new_person._update_organizations()
File "/data/vhost/mark.candidates.dev.mysociety.org/yournextmp-popit/candidates/models.py" in _update_organizations
  91.         for m in self.popit_data['memberships']:

Exception Type: KeyError at /person/stephen-williams/update
Exception Value: 'memberships'

Currently the admin interface drop down selection lists all parties alphabetically. There are hundreds.

Instead it is proposed that the top five parties are listed at the top alphabetically, with the remainder beneath a dotted line:

Conservatives
Green
Labour
Lib Dem
UKIP

You can get Javascript autocompletion of party names directly from PopIt (using JSONP) - the logic for forming the URL would look a bit like this:

• e51c951ad6cc480#diff-168b14845ef8e4c0e0bcf4d65be711a5R580

But it might be helpful to provide a JSONP-accessible API endpoint for this in the front-end as well. This could be helpful for people as well as parties.

This might be useful for simple demographic analysis

Suggested by @symroe

Lots of these are arty shots with lots of landscape, so we'd want to use OpenCV to crop to faces, as SayIt does now.

When inputting a date of birth, I got a 'please enter a valid date' message, but there's no indication of the approved format.

(I'd pasted straight from Wikipedia as "December 22, 1964").

I need to create new Facebook, Google and Twitter keys / secrets for social logins for the live site.

At the moment there's no way to see information about a candidate except by clicking "Edit" beside them (and that's only available to logged-in users). We should add even a very basic page that displays the information we have for that candidate.

A confirmed email address is required even if you're using social login, which is probably excessive. The reason for this at the moment is just not having found a satisfactory combination of django-allauth settings, and wanting to err on the side of discouraging any malicious users at first...

Comments from my earlier email

1. Login with twitter is needlessly complex: we don't need to verify email,
   I don't think. (I've not tried the other methods)

Yeah, it's way over the top, but I couldn't find any other
combination of django-allauth settings that were satisfactory,
and at the moment I'm more worried about junk data being added
by time-wasters than not having enough people take part. [...]
There's some comments about that change to the authentication
here:

c119bff

As I said there, I'm tempted to ditch the social login entirely
(even though I like the idea of storing less user data), partly
to make this simpler and partly because I keep reading things
like this:

https://medium.com/on-coding/the-unexpected-costs-of-third-party-login-cda41c087653

http://candidates.staging.mysociety.org/constituency/Stroud

Add a scraper or scrapers for UKIP candidates, similar to these ones:

• https://github.com/mhl/UK-Political-Parties/tree/ppc-scrapers/ppcs

The "versions" data in PopIt helps you to see what happened to a particular candidate's record, but it doesn't provide an easy view of what particular users have done (and wouldn't record deletion of people, for example). If we get a malicious user, it would also be helpful to be able to go through all the changes they've made to revert them as well. So, a simple "action log" table would help with this.

LinkedIn pages - e.g. https://www.linkedin.com/pub/layla-moran/28/670/284 are an incredibly rich source of background information on candidates. A field should be added to enable these to be populated for the candidate records

We now have a list of registered parties from the Electoral Commission in PopIt so it seems sensible just to offer those as a choice, rather than the current behaviour when anything entered that isn't the same as something in the autocomplete will cause a new party to be created. (This probably would result in lots of typo or alias parties being created.)

The maximum person ID in use should now be set by the candidates_create_popit command, but currently creating new people doesn't use this information. It needs to try the ID MaxPopItIds.get_max_persons_id() + 1 first, increment it until creation succeeds, and then use MaxPopItIds.update_max_persons_id(...) to record the ID that worked.

Tnks

It's confusing having this separate at the moment.

These party logos are in PopIt now (e.g. see https://yournextmp.popit.mysociety.org/organizations/party:52) so we could display them alongside the candidates. (The image proxy should let you request them in a standard thumbnail size, as well, I think.)

Because of the way I imported the data, the image URLs all refer to localhost URLs - I've added a candidates_delete_party_images command and then we can just rerun candidates_update_popit_parties_from_ec_data.