Archive for the ‘api’ Category
Well, that took a little longer than planned…
[I won't go into the details, but suffice to say our internal deadline got squeezed between the combination of a fast-growing website, the usual issues of large datasets, and that tricky business of finding and managing coders who can program in Ruby, get data, and be really good at scraping tricky websites.]
But I’m pleased to say we’ve now well on our way to not just resurrecting PlanningAlerts in a sustainable, scalable way but a whole lot more too.
Where we’re heading: a open database of UK planning applications
First, let’s talk about the end goal. From the beginning, while we wanted to get PlanningAlerts working again – the simplicity of being able to put in your postcode and email address and get alerts about nearby planning applications is both useful and compelling – we also knew that if the service was going to be sustainable, and serve the needs of the wider community we’d need to do a whole lot more.
Particularly with the significant changes in the planning laws and regulations that are being brought in over the next few years, it’s important that everybody – individuals, community groups, NGOs, other websites, even councils – have good and open access to not just the planning applications in their area, but in the surrounding areas too.
In short, we wanted to create the UK’s first open database of planning applications, free for reuse by all.
That meant not just finding when there was a planning application, and where (though that’s really useful), but also capturing all the other data too, and also keep that information updated as the planning application went through the various stages (the original PlanningAlerts just scraped the information once, when it was found on the website, and even then pretty much just got the address and the description).
Of course, were local authorities to publish the information as open data, for example through an API, this would be easy. As it is, with a couple of exceptions, it means an awful lot of scraping, and some pretty clever scraping too, not to mention upgrading the servers and making OpenlyLocal more scalable.
Where we’ve got to
Still, we’ve pretty much overcome these issues and now have hundreds of scrapers working, pulling the information into OpenlyLocal from well over a hundred councils, and now have well over half a million planning applications in there.
There are still some things to be sorted out – some of the council websites seem to shut down for a few hours overnight, meaning they appear to be broken when we visit them, others change URLs without redirecting to the new ones, and still others are just, well, flaky. But we’ve now got to a stage where we can start opening up the data we have, for people to play around with, find issues with, and start to use.
For a start, each planning application has its own permanent URL, and the information is also available as JSON or XML:
There’s also a page for each council, showing the latest planning applications, and the information here is available via the API too:
There’s also a GeoRSS feed for each council too allowing you to keep up to date with the latest planning applications for your council. It also means you can easily create maps or widgets for the council, showing the latest applications of the council.
Finally, Andrew Speakman, who’d coincidentally been doing some great stuff in this area, has joined the team as Planning editor, to help coordinate efforts and liaise with the community (more on this below).
The next main task is to reinstate the original PlanningAlert functionality. That’s our focus now, and we’re about halfway there (and aiming to get the first alerts going out in the next 2-3 weeks).
We’ve also got several more councils and planning application systems to add, and this should bring the number of councils we’ve got on the system to between 150 and 200. This will be an ongoing process, over the next couple of months. There’ll also be some much-overdue design work on OpenlyLocal so that the increased amount of information on there is presented to the user in a more intuitive way – please feel free to contact us if you’re a UX person/designer and want to help out.
We also need to improve the database backend. We’ve been using MySQL exclusively since the start, but MySQL isn’t great at spatial (i.e. geographic) searches, restricting the sort of functionality we can offer. We expect to sort this in a month or so, probably moving to PostGIS, and after that we can start to add more features, finer grained searches, and start to look at making the whole thing sustainable by offering premium services.
We’ll be working too on liaising with councils who want to offer their applications via an API – as the ever pioneering Lichfield council already does – or a nightly data dump. This not only does the right thing in opening up data for all to use, but also means we don’t have to scrape their websites. Lichfield, for example, uses the Idox system, and the web interface for this (which is what you see when you look at a planning application on Lichfield’s website) spreads the application details over 8 different web pages, but the API makes this available on a single URL, reducing the work the server has to do.
Finally, we’re going to be announcing a bounty scheme for the scraper/developer community to write scrapers for those areas that don’t use one of the standard systems. Andrew will be coordinating this, and will be blogging about this sometime in the next week or so (and you can contact him at planning at openlylocal dot com). We’ll also be tweeting progress at @planningalert.
Thanks for your patience.
Since OpenlyLocal started pulling in council spending data, it’s niggled at me that it’s only half the story. Yes, as more and more data is published you’re beginning to get a much clearer idea of who’s paid what. And if councils publish it at a sufficient level of detail and consistently categorised, we’ll have a pretty good idea of what it’s spent on too.
However, useful though that is, that’s like taking a peak at a company’s bank statement and thinking it tells the whole story. Many of the payments relate to goods or services delivered some time in the past, some for things that have not yet been delivered, and there are all sorts of things (depreciation, movements between accounts, accruals for invoices not yet received) that won’t appear on there.
That’s what the council’s accounts are for — you know, those impenetrable things locked up in PDFs in some dusty corner of the council’s website, all sufficiently different from each other to make comparison difficult:
For some time, the holy grail for projects like OpenlyLocal and Where Does My Money Go has been to get the accounts in a standardized form to make comparison easy not just for accountants but for regular people too.
The thing is, such a thing does exist, and it’s sent by councils to central Government (the Department for Communities and Local Government to be precise) for them to use in their own figures. It’s a fairly hellishly complex spreadsheet called the Revenue Outturn form that must be filled in by the council (to get an idea have a look at the template here).
They’re not published anywhere by the DCLG, but they contain no state secrets or sensitive information; it’s just that the procedure being followed is the same one as they’ve always followed, and so they are not published, even after the statistics have been calculated from the data (the Statistics Act apparently prohibit publication until the stats have been published).
So I had an idea: wouldn’t it be great if we could pull the data that’s sitting in all these spreadsheets into a database and so allow comparison between councils’ accounts, thus freeing it from those forgotten corners of government computers.
This would seem to be a project that would be just about simple enough to be doable (though it’s trickier than it seems) and could allow ordinary people to understand their council’s spending in all sorts of ways (particularly if we add some of those sexy Where Does My Money Go visualisations). It could also be useful in ways that we can barely imagine – some of the participatory budget experiments going in on in Redbridge and other councils would be even more useful if the context of similar councils spending was added to the mix.
So how would this be funded. Well, the usual route would be for DCLG or perhaps the one of the Local Government Association bodies such as IDeA to scope out a proposal, involving many hours of meetings, reams of paper, and running up thousands of pounds in costs, even before it’s started.
They’d then put the process out to tender, involving many more thousands in admin, and designed to attract those companies who specialise in tendering for public sector work. Each of those would want to ensure they make a profit, and so would work out how they’re going to do it before quoting, running up their own costs, and inflating the final price.
So here’s part two of my plan, instead going down that route, I’d come up with a proposal that would:
- be a fraction of that cost
- be specified on a single sheet of paper
- paid for only if I delivered
Obviously there’s a clear potential conflict of interest here – I sit on the government’s Local Public Data Panel and am pushing strongly for open data, and also stand to benefit (depending on how good I am at getting the information out of those hundreds of spreadsheets, each with multiple worksheets, and matching the classification systems). The solution to that – I think – is to do the whole thing transparently, hence this blog post.
In a sense, what I’m proposing is that I scope out the project, solving those difficult problems of how to do it, with the bonus of instead of delivering a report, I deliver the project.
Is it a good thing to have all this data imported into a database, and shown not just on a website in a way non-accountants can understand, but also available to be combined with other data in mashups and visualisations? Definitely.
Is it a good deal for the taxpayer, and is this open procurement a useful way of doing things? Well you can read the proposal for yourself here, and I’d be really interested in comments both on the proposal and the novel procurement model.
I had a fantastic response to the launch of OpenCharities — my little side project to open up the Charity Commission’s Register of Charities — from individuals, from organisations representing the third sector, and from charities themselves.
There were also a few questions:
- Could we pull out and expose via the api more info about the charities, especially the financial history?
- How often would OpenCharities be updated and what about new charities added after we’d scraped the register?
- Was there any possibility that we could add additional information from sources other than the Charity Register?
So, over the past week or so, we’ve been busy trying to answer those questions the best we could, mainly by just trying to get on and solve them.
First, additional info. After a terrifically illuminating meeting with Karl and David from NCVO, I had a much better idea of how the charity sector is structured, and what sort of information would be useful to people.
So the first thing I did was to rewrite the scraper and parser to pull in a lot more information, particularly the past 5 years income and spending and, for bigger charities the breakdown of that income and spending. (I also pulled in the remaining charities that had been missed the first time around, including removed charities.) Here’s what the NSPCC’s entry, for example, looks like now:
We are also now getting the list of trustees, and links to the accounts and Summary Information Returns, as there’s all sorts of goodness locked up in those PDFs.
However, while we running through the all these charities, we wondered if any of them had social networking info easily available (i.e. on their front page). It turns out some of the bigger ones did, and so we visited their sites and pulled out that info (it’s fairly easy to look for links for twitter/facebook/youtube etc on a home page). Here’s an example social networking info, again for the NSPCC.
[Incidentally, doing this threw up some errors in the Charity Register, most commonly websites that are listed as http://http://some.charity.org.uk, which in itself shows the benefit of opening up the data. All we need now is a way of communicating that to the Charity Commission.]
We also (after way too many hours wasted messing around with cookies and hidden form fields) figured out how to get the list of charities recently added, with the result that we can check every night for new charities added in the past 24 hours, and add those to the database.
This means not only can we keep OpenCharities up to date, it also means we can offer an RSS feed of the latest charities. And if that’s updated a bit too frequently for you (some days there are over 20 charities added), you can always restrict to a given search term, e.g http://OpenCharities/charities.rss?term=children for those charities with children in the title.
Finally, we’ve been looking at what other datasets we could link with the register, and I thought a good one might be the list of grants given out by the various National Lottery funding bodies (which fortunately had already been scraped by the very talented Julian Todd using ScraperWiki).
Then it was a fairly simple matter of tying together the recipients with the register, and voila, you have something like this:
Note, at the time of writing, the import and match of the data is still going on, but should be finished by the end of today.
We’ll also add some simple functionality to show payments from local councils that’s being published in the local council spending data. The information’s already in the database (and is actually shown on the OpenlyLocal page for the charity); I just haven’t got around to displaying it on OpenCharities yet. Expect that to appear in the next day or so.
A couple of weeks ago Will Perrin and I, along with some feedback from the Local Public Data Panel on which we sit, came up with some guidelines for publishing local spending data. They were a first draft, based on a request by Camden council for some guidance, in light of the announcement that councils will have to start publishing details of spending over £500.
Now I’ve got strong opinions about standards: they should be developed from real world problems, by the people using them and should make life easier, not more difficult. It slightly concerned me that in this case I wasn’t actually using any of the spending data – mainly because I hadn’t got around to adding it in to OpenlyLocal yet.
This week, I remedied this, and pulled in the data from those authorities that had published their local spending data – Windsor & Maidenhead, the GLA and the London Borough of Richmond upon Thames. Now there’s a couple of sites (including Adrian Short’s Armchair Auditor, which focuses on spending categories) already pulling the Windsor & Maidenhead data but as far as I’m aware they don’t include the other two authorities, and this adds a different dimension to things, as you want to be able to compare the suppliers across authorities.
First, a few pages from OpenlyLocal showing how I’ve approached it (bear in mind they’re a very rough first draft, and I’m concentrating on the data rather than the presentation). You can see the biggest suppliers to a council right there on the council’s main page (e.g. Windsor & Maidenhead, GLA, Richmond):
Click through to more info gets you a pagination view of all suppliers (in Windsor & Maidenhead’s case there are over 2800 so far):
Clicking any of these will give you the details for that supplier, including all the payments to them:
And clicking on the amount will give you a page just with the transaction details, so it can be emailed to others
But we’re getting ahead of ourselves. The first job is to import the data from the CSV files into a database and this was where the first problems occurred. Not in the CSV format – which is not a problem, but in the consistency of data.
Take Windsor & Maidenhead (you should just be able to open these files an any spreadsheet program). Looking at each data set in turn and you find that there’s very little consistency – the earliest sets don’t have any dates and aggregate across a whole quarter (but do helpfully have the internal Supplier ID as well as the supplier name). Later sets have the transaction date (although in one the US date format is used, which could catch out those not looking at them manually), but omit supplier ID and cost centre.
On the GLA figures, there’ a similar story, with the type of data and the names used to describe changing seemingly randomly between data sets. Some of the 2009 ones do have transaction dates, but the 2010 one generally don’t, and the supplier field has different names, from Supplier to Supplier Name to Vendor.
This is not to criticise those bodies – it’s difficult to produce consistent data if you’re making the rules up as you go along (and given there weren’t any established precedents that’s what they were doing), and doing much of it by hand. Also, they are doing it first and helping us understand where the problems lie (and where they don’t). In short they are failing forward –getting on with it so they can make mistakes from which they (and crucially others) can learn.
But who are these suppliers?
The bigger problem, as I’ve said before, is being able to identify the suppliers, and this becomes particularly acute when you want to compare across bodies (who may name the same company or body slightly differently). Ideally (as we put in the first draft of the proposals), we would have the company number (when we’re talking about a company, at any rate), but we recognised that many accounts systems simply won’t have this information, and so we do need some information that helps use identify them.
Why do we want to know this information? For the same reason we want any ID (you might as well ask why Companies House issues Company Numbers and requires all companies to put that number on their correspondence) – to positively identify something without messing around with how someone has decided to write the name.
With the help of the excellent Companies Open House I’ve had a go at matching the names to company numbers, but it’s only been partially successful. When it is, you can do things like this (showing spend with other councils on a suppliers’ page):
It’s also going to allow me to pull in other information about the company, from Companies House and elsewhere. For other bodies (i.e. those without a company number), we’re going to have to find another way of identifying them, and that’s next on the list to tackle.
Thoughts on those spending data guidelines
In general I still think they’re fairly good, and most of the shortcomings have been identified in the comments, or emailed to us (we didn’t explicitly state that the data should be available under an open licence such as the one at data.gov.uk, and be definitely should have done). However, adding this data to OpenlyLocal (as well as providing a useful database for the community) has crystalised some thoughts:
- Identification of the bodies is essential, and it think we were right to make this a key point, but it’s likely we will need to have the government provide a lookup table between VAT numbers and Company Numbers.
- Speaking of Government datasets, there’s no way of finding out the ancestry of a company – what its parent company is, what its subsidiaries are, and that’s essential if we’re to properly make use of this information, and similar information released by the government. Companies House bizarrely doesn’t hold this information, but the Office For National Statistics does, and it’s called the Inter Departmental Business Register. Although this contains a lot of information provided in confidence for statistical reasons, the relationships between companies isn’t confidential (it just isn’t stored in one place), so it would be perfectly feasible to release this information.
- We should probably be explicit whether the figures should include VAT (I think the Windsor & Maidenhead ones don’t include it, but the GLA imply that theirs might).
- Categorisation is going to be a tricky one to solve, as can be seen from the raw data for Windsor & Maidenhead – for example the Children’s Services Directorate is written as both Childrens Services & Children’s Services, and it’s not clear how this, or the subcateogries, ties into standard classifications for government spending, making comparison across authorities tricky.
- I wonder what would be the downside to publishing the description details, even, potentially, the invoice itself. It’s probably FOI-able, after all.
As ever, comments welcome, and of course all the data is available through the API under an open licence.
Why was it important that the UK government open up the geographic infrastructure? Because it makes so many location-based things that were tortuous, almost trivial.
Previously, getting open data about your local councillors, given just a postcode, was a tortuous business, requiring multiple calls to different sites. Now, it is easy. Just go to http://openlylocal.com/areas/postcodes/%5Byourpostcodehere%5D and, bingo, you’re done.
You can also just put your postcode in the search box on any OpenlyLocal page to do the same thing. And, obviously, you can also download the data as XML or JSON, and with an open data licence that allows reuse by anybody, even commercial reuse.
There’s still a little bit of tweaking to be done. I need to match up postcodes county electoral divisions, and I’m planning on adding RDF to the data types returned. Finally, it’d be great to show the ward boundaries on a map, but I think that may take a little more work.