{"id":2697,"date":"2020-10-08T15:29:01","date_gmt":"2020-10-08T02:29:01","guid":{"rendered":"http:\/\/blogs.lincoln.ac.nz\/gis\/?p=2697"},"modified":"2023-05-07T03:12:33","modified_gmt":"2023-05-07T03:12:33","slug":"deleting-smokers-part-3-progress","status":"publish","type":"post","link":"https:\/\/blogs.lincoln.ac.nz\/gis\/deleting-smokers-part-3-progress\/","title":{"rendered":"Deleting Smokers – Part 3 – Progress!"},"content":{"rendered":"

In Part 3 of a series on developing a Python script, some progress is detailed.\u00a0 A short data-massaging script helps prepare the input data for the main script.<\/em><\/p>\n

In this ongoing series about developing a Python script for some analysis, we first covered what the script needs to do<\/a>\u00a0and then also covered some data issues<\/a>.\u00a0 After several full on days of banging my head against a digital wall, I think I’ve made a bit of progress!\u00a0 Recall that we’re looking at some analysis that looks at reducing the number of tobacco retailers based on a minimum distance criterion (i.e., there should be no tobacco retailers within a given distance of another retailer).\u00a0 The trick is in systematically removing retailers in a defensible and logical way.<\/p>\n

At the end of our last post, I had the input data down to three options for setting up the proximity relationship: a point layer of retailers that the Near tool has been run on (thus providing distances from each point to the nearest retailer), a similar output from the Generate Near Table tool that is a table rather than a point layer, and the output from an OD Cost Matrix.\u00a0 To be honest, at the start of this I though the cost matrix would be the way to go, but after last week, I think it’s a bit more problematic than it’s worth.<\/p>\n

So I’ve decided to focus on the point layer with the near distances in the table.\u00a0 Some benefits of this are that I’m working directly with the points that will need to be deleted, cutting down on any extra to-ing an fro-ing between other layers or tables.\u00a0 Also, the Near tool allows me to limit the distances that are calculated so I’ll easily be able to know which didn’t have a retailer within that minimum distance (I should be able to do this with the OD Cost Matrix but I haven’t been able to make that work yet).\u00a0 One downside is that I’ll be working with as-the-crow-flies straight line distances rather than distances along the road network (which would be more realistic but at these distances might not make much difference).<\/p>\n

So, here’s what my input data table looks like after running the Near tool:<\/p>\n

\"\"<\/a><\/p>\n

Each point has an OBJECTID, a unique identifier that Pro uses to keep track of each point.\u00a0 The NEAR_FID is the OBJECTID of the nearest point and NEAR_DIST is the straight line distance between the two.\u00a0 So the point with OBJECTID = 1 is 143.898713 metres away\u00a0from the point with\u00a0NEAR_FID = 274, for example.\u00a0 When running this tool, I set the search distance to 500 m, meaning that if it doesn’t find another point within that distance, it will return a value of -1 for both NEAR_FID and NEAR_DIST.\u00a0 You can see two examples of that in the table above.<\/p>\n

In the last post I was rattling on about reciprocals and needing to link them.\u00a0 Here’s what I mean – if I sort the table by NEAR_DIST and scroll down past all the -1s, you can see that for every OBJECTID\/NEAR_FID distance, there’s its opposite (reciprocal), where the OBJECTID and NEAR_FID are swapped:<\/p>\n

\"\"<\/a><\/p>\n

I’ve highlighted one pair to illustrate – each point is listed separately but they have the same distance.\u00a0 This is a good thing as this is the proximity relationship between these two points.\u00a0 Later in the script I’ll need to delete one of these – but first I need to know which points are these de facto relationships.<\/p>\n

(While we’re here, notice that there are some points that don’t have reciprocals – at first these were keeping me awake at night but I’m now a lot less worried about them.\u00a0 More on these later.)<\/em><\/p>\n

Okay – so while this allows me to know the proximity relationships, it’s a bit too complicated at this stage to work with these, so I’m wanting to do a bit of data massaging, for which I’ll use a wee script to identify those points that have reciprocals and assign them a unique id – I can then later use\u00a0this in the main script as my proximity relationship identifier (did that statement make any sense at all?\u00a0 (Ed. no, not really, but carry on…<\/em>)<\/p>\n

Naively, I thought I’ll just whip up this script in a few minutes and move on.\u00a0 That’s when the banging me head up against a digital wall started – and boy, does my aura hurt.<\/p>\n

Here’s the guts of what I wanted to script to do (psuedo-code):<\/p>\n