Moriarty Naps

← back

Command Carto II – Make

Work in progress, tread lightly

Welcome back!

Last time we popped open our terminal and did some quick GIS work. As we saw at the end, the real benefit comes from chaining commands together. However, typing all that out in the terminal is inconvenient, and you don’t leave a history of commands for later.

We want to yell “DO THE THING”, in so many words, and have it do the thing.

Good news! This round we’re gonna learn how to do that by introducing Makefiles.


Making Camp

At this point I’d recommend getting a text editor to help write in. Text editors are fancy notepads that make coding more pleasant.

I personally use Sublime, but there are lots of other good options – all free to download and use. If you work on a team, it’s best to see what your colleagues are using and start there.


  1. New blog post, new package of files to help ya along. Download this zip file (409kb) s'il vous plaît.
  2. Throw those files into a folder, and navigate your terminal to it.
  3. Make sure your terminal can run Make commands. You can test this by typing make into your terminal. If it returns the message: make: *** No targets specificed..., you're good. Otherwise, you may need to install additional developer tools.
  4. Mentally picture your favorite meal or snack. Promise yourself that you'll go get said meal or snack after finishing this tutorial. Makefile syntax is frustrating and unforgiving. You'll have earned it.


Make It So

Make was created in 1976 as a means to automate commands for setting up programs. It’s still used, albeit mostly in software development.

Instead of writing commands in the terminal one by one, we’re saving them to a file that we can call from instead. That file is a makefile.

In our Makefile, we write Tasks In official documentation, this is called TARGET. I’ll explain why in the next installment. For now, don’t worry about it. . Think of each Task as a lil’ robot you can call to execute your series of commands. A Make task is formatted like so:

To call a task, you first save your Makefile to a folder. For brevity just calling it makefile is fine. If your terminal is directed to a folder containing a Makefile, you can call tasks from the terminal with: make [task name]. When called the commands for that task are run sequentially.

Mike… Alfa… Kilo… Echo…

We’re going to be alternating between your terminal and the Makefile throughout this tutorial. For the sake of clarity, code written in the purple boxes is meant for a Makefile. Code in white boxes are meant to be entered into your terminal, just like last time.

The following is a functional make command:

echo-caynon: echo "don't have a cow homer!" echo "don't have a cow homer!!" echo "don't have a cow homer!!!"

The echo-caynon task is included in the Makefile you downloaded. You can run it yourself by typing make echo-caynon in your terminal and hitting enter. Behold! It executes each line.

echo is a new command. It prints out in your console whatever follows the command.

Familiar Functions

With this format, we can bake in the commands from our previous work into a task we can call in three words:

pandaMap: mapshaper countries.json \ -join native-red-pandas.csv keys=NAME,COUNTRY \ -filter '"yes".indexOf(PANDAS) > -1' \ -clean \ -proj +init=EPSG:32645 \ -o countries-with-pandas.json echo "finished"

This is also included, so run make pandaMap and sit back and enjoy watching your Makebot do all the work for you.

The \ become more important here. Without it each new line is interpreted as a new command, as you saw in echo-caynon. \ tells the Makebot that the next line isn’t new, but rather a continuation of the previous lines command. Everything will work if you typed it out one uninterrupted line, but this makes our code more readable. Your future self will thank present you.


Before we get to the meat of this post – a few formatting quirks.

You can have as many tasks as ya wish in one Makefile. They just need to have unique names.

When Make fails, it is very bad about explaining what went wrong. If you get the vague error message: makefile:25: *** missing separator. Stop., you should check to make sure you used tab indentions instead of spaces.

If that’s not the issue (it will be 90% of the time), I’ve found the fastest way to debug is to:

  • Temporarily undo the latest work to determine what new code is introducing the errors.
  • Test out new commands in one-off tasks before baking them in with other commands.
  • Make sure you saved! If you haven’t saved your changes, the file won’t have your updates in it.

K. Let’s GO!


Make a Map

We want to make a map of all the burn footprints from California wildfires for 2018. Our makefile will handle:

  • Download the fire perimeter data
  • Filter the features
  • Dissolve the features
  • & Clip the data to California

If you haven’t already, open “Makefile” from the downloaded resources in your text editor.

From this point on, none of the following tasks are in the included Makefile. You will need to add them to the provided Makefile.

The commands should work as written. But if you need to double check against one that’s pre-written in full, you can find that here.

Fetching the Data

The USGS does a great job providing up-to-date shapefiles for each and every wildfire perimeter as it develops. They also have a file that includes every perimeter shape for the year to date, which we could go get ourselves…

…but we’re being lazy. Let’s tell our bot to go fetch it instead!

curl -O URL
stands for → Client URL

USGS hosts their data on a server called GeoMAC. curl will look at whatever URL you put after the -O, and download it.

fetch: curl -O ""

Curl-ing has the nice feature of displaying how the download is going in text. A lot of commands don’t give an indication of how long they’re taking to complete. In general, you’ll know a command is done when your cursor returns in the terminal.

Unzipping the .zip file

If you run ls in your terminal, you should now see the zip file listed. Let’s get those files out. But, one brief detour first –

Is the first statement true? If not than do second part.

Unzipping this file is going to spit out a pile of shapefile files. Let’s keep things tidy by putting them into their own folder. This command tests if a folder exists, and if it doesn’t, it executes the mkdir command on the rightside of the ||.

”||” is a logic operator, but don’t fret. We’ll only be using it to using it to create folders.

unzip -o "FILE" -d DESTINATION

This will unzip the file. The text following -d will tell it where to save the file.

You’ll often see dash & letters following commands. They are shorthand for longer words, cut down for brevity. For most CLI libaries you can type -h to get a list of what those commands do. That’s your Try it with unzip.

unzip -h

Put together those commands into a Makebot:

unzip: @[ -d data ] || mkdir data unzip -o "" -d data/

To review, our first command checks to see if a folder named ‘data’ exists and if not makes one. Our second command unzips the zip file, and puts it in our ‘data’ folder. Spiffy.


The downloaded file has the perimeter from every wildfire in the U.S. for 2018. We don’t need info for what was happening on Florida, so we can filter out all states that aren’t California. Here’s a filter task:

filter: mapshaper data/2018_perimeters_dd83.shp \ -filter '"CA".indexOf(state) > -1' \ -o data/perimeters_filtered.shp

It may take a second or a minute, but you should see a message “[filter] Retained 992 of 6,046 features” after running make filter.

You may notice that both our input and output have a data/ in front of the filename. That tells our terminal to peek into the data folder.

Filtering helps cut down the filesize considerably, but the output still contains multiple perimeters for each fire (they measure it daily). For our purposes we only want the whole burn area for each fire, so let’s dissolve the features.


mapshaper INPUT -dissolve2 FIELD -o OUTPUT

The ‘FIELD’ is what we’re choosing to dissolve the features by. In this case the field “incidentna” (incident name) should do the trick.

Translated into a Makebot task:

dissolve: mapshaper data/perimeters_filtered.shp \ -dissolve2 incidentna \ -o data/perimeters_dissolved.shp

Run make dissolve and now we have the overall fire perimeters.

We’re still not done, however, as some of these fire perimeters cross state borders. We don’t need that cross state information, so let’s clip it out.


Mapshaper INPUT -clip CLIPBY -o OUTPUT

This will clip your INPUT file by the CLIPBY file and save it as OUTPUT.

I provided a clipped version of California in the download package. We’ll use that as our CLIPBY file.

clip: mapshaper data/perimeters_dissolved.shp \ -clip California.geojson \ -o perimeters_clipped.shp

Run make clip and we’re good. to. go.

You could overwrite the original file for these if you choose an OUTPUT that is the same as the INPUT. In practice, it’s much better to make a new file for every operation. This will help you debug if there’s an issue, and allows you to always be able to run your tasks and get the same results.


Bringing it altogether

You can call Make tasks from other tasks in your Makefile! This lets us keep our tasks piecemeal, as we’ve been doing, while still letting everything get called from a sole call.

Once everything’s working we can call them all through a “factory” This is not official terminology. Just how I refer to these sorts of tasks. task.

wildfire-factory: make clean make fetch make unzip make filter make dissolve make clip

Alas, I did sneak one more task into that.

Cleanin’ Up

clean: rm -rf data/ rm -f

rm is the standard CLI operation to ‘remove’ a the file that follows. -f and -rf are ways to ‘force’ it, which will ignore non-existent files.

This deletes the zip file and our data folder. This can be helpful for a few reasons: you save space on your computer, and you’ll know that your factory will run under the same conditions next time.

When you’re using these commands, always explicitly state what you want deleted. -rf in particular can be dangerous if it’s used in a root folder with a catchall * or “.”. We can avert that danger by:

  • Always stating exactly what file/folder you want removed, no catchalls
  • Only remove files you’ve fetched or created through your makefile. Things you can always recreate if need be.

Add the above tasks to your file and run the make wildfire-factory task. Viola!

If something goes awry with one of the tasks, you can always troubleshoot it by calling the problem issue on it’s own. Heck, you don’t have to pile everything together either. If you don’t want to fetch every time, for example, just keep that task separate.

Do we need to write each mapshaper command in it’s own task? Nope. These could also be written in one straight command:

wildfire-factory: rm -rf data/ rm -r rm curl -O "" @[ -d data ] || mkdir data unzip -o "" -d data/ Mapshaper data/2018_perimeters_dd83.shp \ -filter '"CA".indexOf(state) > -1' \ -clip California.geojson \ -dissolve2 incidentna \ -o format=geojson data/perimeters_filtered.json

That said, splitting the tasks up makes for easier testing while working things out. It’s also more human readable. When you write these yourself, it’s entirely your call on how you organize things.

The Real Magic

Let’s say you’re so excited by this newfound power that you want to see what the 2017 wildfire situation looked like.

All you have to do is replace every instance of 2018 with 2017 in your Makefile, then run your factory command. Try it out! Be amazed.

This format can take awhile to get used to – I still troubleshoot Google frequently. The good news is that you now have working commands for doing basic GIS operations. If you come across a job later that you need a CLIP task for, you already have a working blueprint.

The world is yours to takeover with robots.


ALL the Buildings

Before we go, let’s build one more bot.

In 2017, Microsoft released a dataset with all of the buildings in the United States. They have it free for you to download (nice!), but only as geojsons. That’s a lovely format but when you’re working with 2gb geojsons QGIS will start hemmoraging. Let’s make a bot that downloads the file for us and makes a quick conversion to the easier-to-parse shapefile.

First thing first, we’ll want to make sure there’s a folder for the data.

getFootprints: @[ -d data ] || mkdir data

The URL we target for downloading is

This is perfect for us, because the URL is based on the state we want. If you change Wisconsin there to Idaho you’ll hit a file for Idaho instead. This consistency is great.


In a Makefile, there are two places you can create variables. The first place is typed in the document itself, along the lines of:


Where VARIABLE can be any ol’ declarative you want to use. For Makefiles, I tend you write my variables in UPPERCASE to help keep them visually distinct. When it comes time to write a bot, you can now reference that variable via:

testbot: echo $(VARIABLE)

The $() bit tells Make that the name inside is a variable it should be looking for, and not just text (or a command) that reads VARIABLE.

The second place to assign a value to a variable is when we call the command. Delete the line in your Makefile that assigned “value” to VARIABLE.

Now run the following command in your terminal…

make testbot VARIABLE="different value"

…and our testbot task will echo that value!

We can take advantage of this and Microsoft’s fortunate URL structure to make our fetch code grab a STATE we choose when we call the task. In this case, we’ll curl:$(STATE).zip.

getFootprints: @[ -d data ] || mkdir data curl -O$(STATE).zip data

Try running that with the state of your choice that isn’t California, Florida, Ohio or Texas! You should be rewarded with a downloaded file. I’ll explain about California below.

make getFootprints STATE=Wisconsin


That grabs the zip file. Now let’s add an unzip command to the task:

getFootprints: @[ -d data ] || mkdir data curl -O$(STATE).zip data unzip data/$(STATE).zip -d data

Because we’re still in the same task, $(STATE) will still be understood as whichever value you called with.

Let’s convert it a shapefile. If you run mapshaper on it’s lone self, you may get an error: “FATAL ERROR: CALL_AND_RETRY_LAST Allocation failed — process out of memory”.

These files are huge, which means that we need to allocate more computin’ power to Mapshaper to finish the task. --max-old-space-size=8192 `which mapshaper` will do this.

getFootprints: @[ -d data ] || mkdir data curl -O$(STATE).zip data unzip data/$(STATE).zip -d data node --max-old-space-size=8192 `which mapshaper` data/$(STATE).geojson -o format=shapefile data/$(STATE).shp

If you run this, it may take five to ten minutes, but during that time you can go do other things. Make breakfast. Do some jumping jacks. Make yourself a fancy cocktail.

Cleaning House

Finally, given that we only wanted the Shapefile, it can be nice to remove some of the other files we gathered on the way.

getFootprints: @[ -d data ] || mkdir data curl -O$(STATE).zip data unzip data/$(STATE).zip -d data node --max-old-space-size=8192 `which mapshaper` data/$(STATE).geojson -o format=shapefile data/$(STATE).shp rm data/$(STATE).geojson

Try that out with a different state! Should work well. Now you can quickly pull down that building data and have it in an already accessible format to clip from.

Texas, California; Taming the Beast

Finally, I mentioned above to not download some of the largest states. That’s because even with the increase in processing power, Mapshaper may not be able to fully convert those files.

Luckily, with CLI GIS, there’s always at least three different libraries to do common operations. In this case, we can use Ogr2Ogr to make the conversion instead– and this one works.

If you have QGIS installed, you already have Ogr2Ogr. However, if you haven’t and just want it you can download it along with the rest of GDAL here.

largerFile: ogr2ogr -nlt POLYGON -skipfailures $(STATE).shp $(STATE).geojson OGRGeoJSON


That's All For Now

Way to go! Writing terminal commands this way can be a real struggle bus. I hope you got some good stuff out of this, and have some ideas for processes ya want to codify in a makebot.

This is just scratching the surface on what Makefiles can do! In the next chapter, I’ll go over how to write more generalized tasks and loops. Stay tuned.

– D.M., make bed LOCATION="brooklyn"

If any parts of this seem super confusing, or you just have any other questions feel free to send a bird my way on Twitter.

Once again, all the kudos to Matthew Bloch for his continued work on Mapshaper.

Also to Seth Fitzsimmons and Joshua Stevens who both gave talks on Make during PCD over the past two years. Joshua's talk does a great job of highlighting the potential strength of working directly with CLI libaries. Meanwhile Seth's talk is a blitz of shell commands that'll make your head spin. After working with make for a bit, you'll undoubtedly pick up some terribly useful code.

And finally a quick kudos to Derek Lieu for introducing me to this approach for GIS work back in 2016.

← Home