Overcoming the Make Query Table bug in ArcGIS

According to my notes, I first used the Make Query Table tool in my first week at Aurecon, back in March 2012. It was the first of many, many times, because often when receiving spatial data in a non-spatial format from a non-GIS user, the first thing that gets thrown out is any trace of the original spatial component.

At some point, I realised the tool’s expression parameter was a bit wonky. As I have come up against this problem every few months since (forgetting it happens each time because I only thought to write down a note about it now), I have decided to immortalise it in a gist below.

Using a query table to represent a 1:M relationship spatially

I first discovered (and used) the Make Query Table tool during my second week at Aurecon (March 13 2012, according to my OneNote). This was about a month before I started using ModelBuilder (to combat the frustrations of ArcMap), and about 6 months before I started Python (to combat the frustrations of ModelBuilder).

I’m just giving a bit of context because this was before everything clicked into place for me. Before this point, I treated everything I learnt in my studies and during my internship as separate silos of information: GIS, Databases, Programming.

I did not even realise until after I was working at Aurecon that the query expressions used in ArcMap are SQL, despite me studying all those things. It just shows how one’s mindset can block progress, and how I allowed the awful experience of learning Computer Science in Afrikaans to stop me from letting things “click” for so long.

Back to the tool. Joins in ArcGIS are notoriously slow, and 1:M joins are not allowed (technically they are, but in the sense that an arbitrary matching feature will be joined). Naturally, the relationship between the GIS and the asset register is 1:M.

For example, a single point is used to represent the physical, spatial location of an asset, reservoir WRV-00001. In the asset register, this reservoir is unbundled into various components – storage tank, building, fence etc. Each of these assets have their own unique asset ID, but all have the same GIS ID.

I now need to represent all these assets spatially. The points/lines will all be on top of each other, but that’s fine. The Make Query Table tool does exactly this, but it is…quirky. I’ve compiled a list of things to remember when using this tool (supplemented by this question on GIS.SE):

  1. Tables/feature classes in the relationship should be stored within the same database: I tend to remember this step only after I add the inputs and the tool shouts at me.
  2. Add the feature class first in the multivalue parameter: The format of the input relies on the format of the first argument in the multivalue parameter control. The feature class should be added first to ensure that the output is a layer, otherwise it will be a table view.
  3. Enclose table names and field names in quotes: For example, I wish to join the asset register table ar to the point layer points using their common field GIS_ID. By default, the tool encloses my whole expression in quotes "points.GIS_ID" = "ar.GIS_ID". This will cause the tool to fail. Add extra quotes around the field names "points"."GIS_ID" = "ar"."GIS_ID".
  4. Choose the NO_KEY_FIELD option: Trying to add key fields causes some erratic behaviour (???). Just don’t do it. By selecting this option, the existing ObjectID field from the input will be used.
  5. The output layer will appear to have no symbology: Go into the Layer properties, click the Symbology tab, click the existing symbol, then OK, OK. It’s a bug./li>
  6. Persist to disk!: Remember to export the layer to a feature class, otherwise the layer will only exist in the map document.

Accessing metadata using ArcPy, Python, and now hermes!

Two weeks ago, a colleague asked me to write a script to extract some metadata values from dozens of feature classes in a gdb, and write it out to a spreadsheet along with some other descriptive properties. I fiddled around with the metadata using Python’s xml package, and managed to come up with a script for her.

So it was to my immense delight that this post popped up in my feedly on Friday. I immediately starred hermes on GitHub, and maybe this will be the first project I can actually contribute to, and not only because of the Futurama reference.

Developing an asset management GIS data maintenance methodology: Part 5 – My preparation for the way forward

(This is Part 5 of a week long series of posts on the current project I am working on, to develop a reasonable GIS data maintenance strategy for Asset Management data. Read Part 1, 2, 3, 4.)

The work I’ve done over the last few weeks (and years) is all leading to one point. A single system, with all the data topologically correct, standardised and easily accessible.

An enterprise geodatabase, with versioning enabled for the Desktop team to maintain the data without fear of conflicts. Using SDE in this manner will automatically allow for checks to be in place, for example, where I could first check the reconciled versions before moving it to default.

Archiving would be set up, and the database would be regularly backed up. Map services would be published and be made available to the non-GIS users such as the Asset team. I’d prepare the services for consumption in their app of choice: ESRI Maps for SharePoint, ArcGIS for AutoCAD, a JavaScript viewer, an Excel document with the attribute tables embedded…

The services would have the sync capability enabled, so that when they go out in the field, a Collector map could be easily configured for data capture in an offline environment. Since they visit areas which are routinely out of cell coverage, this would be ideal (and better than carrying around a printed mapbook).

While I am busy with this year’s updates, I am keeping all of this in mind. Every little bit I can do now is a little bit less that I have to do later. Once this foundation is in place, we can start looking at more advanced aspects, such as turning all this data into geometric networks, and creating custom tools which can automatically calculate the remaining useful life, asset depreciation and answer all the questions that could possibly be asked.

Developing an asset management GIS data maintenance methodology: Part 4 – Data handover and the audit process

(This is Part 4 of a week long series of posts on the current project I am working on, to develop a reasonable GIS data maintenance strategy for Asset Management data. Read Part 1, Part 2 and Part 3 here).

Once all the new data has been processed and captured, or existing data flagged as asset upgrades, we have to extract several attributes for use by the asset guys. This includes the all important GIS ID, which will be used to link the asset register back to the GIS. The asset register contains a mountain of information about the status of the assets, remaining useful life and other financial information. In the GIS data, we are purely concerned with geometry: is this asset in its correct physical location, with the correct dimensions?

We will carry a few other attributes, such as material type, diameter of pipe, name (if relevant) and town. Once I’ve compiled all the data into this structure, I convert it to a spreadsheet filtered by feature type and including the lengths of the line data. The GIS length is used over the length supplied by the client as it represents the situation on the ground.

Once everything has been submitted to the client, a short while later the auditors will appear. Using some algorithm, they select a sample of the assets and send us a list of GIS IDs or asset names. I link this list up to our GIS and provide a KML file of the locations. This proves that we know where those assets listed on the asset register are (and that no one is trying to fabricate assets). Sometimes the auditors will only provide the asset names – that makes it a bit harder for me to link since we don’t always have the name attribute. It’s the reason why the GIS ID should appear in all documentation.

For one client, I went along with the project leader and sat with an auditor in front of ArcMap. I had to physically look up the assets as she read out the GIS ID or asset name. At that time, the data was not in a good state to begin with, so fortunately she had only sampled assets that I could easily retrieve. As part of this methodology development, I’m documenting the state of the data as it was received, the problems it causes (including unnecessary delays) and the estimated time it would take to clean up the full database for each client.

That’s the part I’m most excited about in this process. I can finally get my hands into the data and standardise it, so if there are any auditor queries, or someone wants to know something quickly about a particular asset, I will know exactly where to find it.

Developing an asset management GIS data maintenance methodology: Part 3 – Data (Processing, updating, working the magic)

(This is Part 3 of a week long series of posts on the current project I am working on, to develop a reasonable GIS data maintenance strategy for Asset Management data. Read Part 1 and Part 2 here.)

GIS data. ESRI has a tab in their help documentation dedicated to the topic. As a GIS person, I may have become accustomed to receiving (and accepting) data in almost every conceivable format, because I know that by some process, I can get it into GIS.

Here are a few examples of data I have received over the last few weeks, and the conversion process followed for each:

  • As-built drawings (dwg/dxf), coordinated, but missing a constant: Set the coordinate system in ArcCatalog. Add the study area boundary to ArcMap, followed by the Polyline layer from the CAD file. Right click the Polyline layer > Properties > Transformations > Enable Transformations > Coordinates. Click OK and go back into the Properties to enable transformations again because ArcMap only remembers it the second time. Add the constant to the coordinates, and press OK. Hope that the drawing falls in the right place. When it does, convert the Polyline layer to a feature class. Project it, then inspect the attribute table. Hope that the CAD technician has placed matching elements in the same layer. Extract the features by unique attribute, ignoring any extraneous data of type Circle/Arc, or any layout items.
  • Dozens of KML files, each containing a single point: Batch convert the KML files to feature classes. While this does create multiple GDBs, it ensures that each file is checked before being extracted for processing. If I’m fairly confident in the data, I will run a script to convert the KMLs and merge them into a single feature class at the same time.
  • PDF files of maps which have been drawn on and scanned: Convert the relevant pages from the PDF to JPG. Georeference the JPG if it contains identifying features such as cadastre (with erf numbers). If there is no line data, it may be OK to eyeball it when digitising.
  • A0 hard copies of maps/as-builts with no digital copy: Eyeball it.
  • Spreadsheets with road names and intersections (no GIS IDs): Format the spreadsheet so that it can be converted (no spaces in field names, remove unnecessary columns etc) to a file gdb table. Hope that the unique combination of road names and intersections will match perfectly with the road feature class.

Sometimes the CAD files are not coordinated, in which case I send it back. Sometimes we get old shapefiles, which have long lost their unique GIS IDs. One time, I received a personal geodatabase (!!!) containing feature classes with a single ID attribute each. Their “matching” attribute tables were stored in separated dbf (!!!) files per folder per service. These dbfs contained many attributes, everything besides the IDs needed to join the data back to the shapes. This is where I had to “work the magic” to get anything usable out.

I haven’t covered all the scenarios, but that’s just about getting the data into GIS. Once it’s in, the data needs to be digitised (if it’s new assets which have been added), or the previous datasets must be inspected and the relevant features extracted (if an asset has been upgraded).

Some discretion needs to be used throughout this process. Time constraints and the current state of the data for a municipality will determine the level of detail which is captured. By putting this methodology in place, I am hoping to change that approach so that in the future, a standard amount of data is captured in a standardised way.

Due to the growing amount of features we were being asked to record as assets, I decided to create a spreadsheet (which will eventually become a table in a database) to separate the services and to specify the prefixes needed for the GIS IDs. A GIS ID is composed of a prefix, a dash and a string of numbers. For example, the prefix for Water Reticulation Pipeline is WRP, so a feature in this feature class may have the GIS ID WRP-00101. Whenever a new asset is added, I run a script to autopopulate the next GIS IDs.

Currently, the list contains over 70 feature types we need to maintain. Each service has its own feature dataset. For example, fds_WaterSupply contains ftr_wrp_Pipe, ftr_why_Hydrant, ftr_wva_Valve etc. The naming convention is not only for consistency (all Water Supply prefixes start with W, all Stormwater prefixes start with SW), but also for the eventual transition to a SQL Server database. This way, the feature classes will be grouped according to the service it belongs to (because SQL Server Management Studio will display the feature class tables in alphabetical order, and because it ignores feature datasets. One of our clients actually pointed out this helpful tip).

I am enforcing certain topology rules based on the requests of the asset guys, such as roads are captured intersection to intersection, sewer pipelines are captured manhole to manhole, water pipelines are captured road intersection to road intersection, and parking areas are captured as polygons and converted to centroids with the polygon area attached to the point. I don’t have the actual topology set up, because at this stage it would add unnecessary complexity. Rather, this capturing convention will become a habit, and as we clean up the older datasets, we will automatically be cleaning topology errors as well.

Despite speaking at length about the data, I have only scratched the surface of what we do with the asset data when it comes to us. In Part 4, I will talk about what happens to the data once it’s been processed by GIS.

Developing an asset management GIS data maintenance methodology: Part 2 – Designing the workflow

(This is Part 2 of a week long series of posts on the current project I am working on, to develop a reasonable GIS data maintenance strategy for Asset Management data. Read Part 1 here.)

Yesterday, I gave a very brief overview of what it maintaining asset management data in GIS involves. Of course, the reality is much more complicated. Over the years that I have been involved with this process, my role has grown from simple data capture, to having full control of the data and the workflow needed to maintain it.

As the amount of work this year increased dramatically, I realised that I would need to document the process that I’ve had in my head. The workload would need to be shared with the Desktop GIS team. The first thing I did was to create a coherent folder structure for data storage. While we are transitioning the system, all of our data is still file-based. This already poses a challenge, as multiple people would need to access the data, and would most likely keep copies on their own devices to avoid data loss/slow access times on the network.

This is what I came up with:


and this is the script which creates that basic structure:

It is still my intention to optimise that script and integrate it into my Asset Management Python Toolbox, but I haven’t had the time do it yet. That structure may look like overkill, but it’s the way I keep my sanity, and allows me to build checks into this process. If at any point the data needs to be reverted, I can retrieve the previous version. For example, if we receive a CAD drawing, the workflow is as follows:

  1. Save the CAD file under data\recd\date_assetperson
  2. Make a copy of the CAD file under conversion\cad_gis\date
  3. Georeference the CAD file using the projection of the local municipality/study area
  4. Convert the CAD layers (normally only the polylines, sometimes the points) to feature classes in a file gdb in the same folder
  5. Project the converted feature classes to a file gdbworkspace\date
  6. Process the data further in that working gdb
  7. Final working data is pulled from all the working gdbs and sorted in current.gdb
  8. Final data for handover is pushed into Municipality.gdb

This process has already saved me time over the last few weeks, where I had to fall back on a previous version of a feature class taken from a CAD drawing due to changing requirements. This is also why I am designing this workflow in an agile way – the requirements are constantly changing, and the data is different for each municipality. I’ve had to add more folders/feature datasets since I drew this up a month ago, and I’m still ironing out the kinks in the communication with the rest of the team.

That brings me to the next aspect of this workflow: the OneNote Notebook. The Asset GIS Notebook contains an Overview section at the top level, which has the contact details of the GIS team members and Asset Management team members. It also contains a breakdown of the folder structure with detailed explanations, as well as links to relevant external documentation, such as the Prefixes spreadsheet (more about that in Part 3).

For each municipal/study area folder in the Assets folder, there is a corresponding section group in the Notebook. This section group contains a General section (technical information such as projection, major towns in the region, project costing details) as well as sections per year. The year section contains all the tasks for the current year, such as

  1. Convert data received
  2. Create mapbooks for site
  3. Generate report

etc. Some of the tasks will be quite specific, depending on the state of the data and the client requirements. There is also a Paper Trail subsection, for all email/Lync/phone correspondence with the asset team. Any answers to questions we have about the data are recorded in this section, not only to cover the GIS Team for audit purposes, but also in case a team member must pick up a task where I have left off.

Of course, it would be terrible to lose all of this hard work. In lieu of a better system, I have MacGyvered a backup system of sorts, where each day before I leave, I sync the network folder to my local OneDrive for Business folder using SyncToy, which then syncs to the cloud. It’s not ideal, but it’s better than what I had before (which was nothing. If that network drive failed…)

Although there are other team members helping with the data capture portion of the work, and who have contributed to the development of the workflow, I still retain responsibility for the process. After they have finalised their data capture, I check the data in the feature classes, assign new GIS IDs (another tool for the toolbox) and load the data into its final structure (also another tool for the toolbox).

Tomorrow in Part 3, I will talk about the kind of data we work with. It will probably be even longer than this post, and will hopefully shed some light on why designing this workflow has been challenging.