GeoJSON

148

ttosh about 17 hours ago 71 commentsRead Article on geojson.org

⚡ Community Insights

Discussion Sentiment

77% Positive

Analyzed from 2572 words in the discussion.

Discussion (71 Comments)Read Original on HackerNews

Stratoscope•about 13 hours ago

One task where GeoJSON falls down is simplification of a group of polygons with common boundaries, e.g. the 48 conterminous US states. If you start with a highly detailed set of polygons, you need to simplify them for practical display in an online map.

GeoJSON doesn't encode the fact that the boundary points are common between adjacent polygons. When you simplify those polygons, each one is handled separately, and you end up with "slivers" where the boundaries are misaligned:

https://www.bing.com/images/search?q=map+slivers+betwen+poly...

TopoJSON solves this by encoding each such boundary only once. So when you simplify the polygons, they are all done together, and the same simplification applies to adjacent polygons. No more slivers!

https://github.com/topojson/topojson

https://github.com/topojson/topojson-simplify

Demiurge•about 13 hours ago

Is this actually GeoJSON falling down, or decades of convention extended to JSON? Topology is great, but it is sidestepped by Shapefile/WKT/WKB/etc, in favor of independent primitives like POINT, LINE, POLYGON. If GeoJSON did not exist as a new JSON GIS data format encoding these primitives, TopoJSON would not have "replaced" it, due to the added mis-match with other non-topological formats.

From what I can tell, the top criticism of GeoJSON is the under-enforced winding order specification, and crossing the antemeridian.

jvanderbot•about 13 hours ago

Right. Encoding a union algorithm into the data structure just introduces the reverse problem: Selecting a subset now requires extra logic beyond jq.

Stratoscope•about 8 hours ago

Similarly, typical map APIs like the Google Maps API accept GeoJSON and not TopoJSON. I was not suggesting TopoJSON as a replacement for GeoJSON, but as a complement to it. With the tools on the TopoJSON GitHub, you can have GeoJSON input and output, but convert to TopoJSON for the simplification step to avoid the "slivers" problem.

pramsey•about 11 hours ago

GeoJSON is not TopoJSON. Saying that is "falling down" is like criticizing a zebra for not being a giraffe. GeoJSON is a mapping of the (non-topological) "simple features" model into JSON, full stop. It does that fine.

Stratoscope•about 8 hours ago

Yes, the same "slivers" problem occurs when you try to simplify features in any format that uses individual polygons, such as shapefiles or whatnot. That's the only case I was referring to.

I don't think I would trust a zebra or a giraffe for this task either.

echoangle•about 12 hours ago

How is that a geojson problem? If your dataset is correct, adjacent borders will just use the same points and will match exactly.

sdenton4•about 11 hours ago

The problem is simplification. Suppose two regions share a border with some nonlinear points a, b, c, d. Simplifying the polygon for the first region might yield a, b, d while the second yield a, c, d. This creates gaps or overlaps between the two regions.

qurren•about 10 hours ago

But what is the border? Set the border to what it actually is, not a simplification of it. The state of Colorado is formally a 697 sided polygon, don't simplify it to a rectangle.

echoangle•about 10 hours ago

So don’t simplify the shapes on their own. Geojson is a storage and exchange format, you can still convert it to other formats if you want to modify it.

NelsonMinar•about 9 hours ago

I like TopoJSON and have used it in projects. But it's weird to set it up as opposition to GeoJSON. It's a complement. GeoJSON is a general data format meant to replace uses of ESRI Shapefiles and other complex formats. TopoJSON is more of a solution for a particular application need.

Is there much work developing or using TopoJSON these days? I haven't seen much about it in a few years.

Stratoscope•about 8 hours ago

To be clear, I'm not suggesting TopoJSON as an alternative to GeoJSON. I like GeoJSON and was loosely involved with the working group that created and updated its spec.

I'm just saying that for the specific task I mentioned GeoJSON or any format such as shapefiles that store polygons individually naturally leads to the "sliver" problem.

A nice processing pipeline is:

1. Convert GeoJSON to TopoJSON.

2. Run the simplification on the TopoJSON.

3. Convert the resulting TopoJSON back to GeoJSON.

The TopoJSON GitHub has tools for each of these steps.

Waterluvian•about 14 hours ago

I’ve applied GeoJSON (among many other GIS tech) for mapping and monitoring tens of thousands of warehouse robots. It works great as long as you squint just a bit, ignoring that it generally calls for long,lat and is designed with the assumption of a world CRS.

The dangerous part is that some tools fully assume this and will completely screw with calculations if you’re assuming a flatland CRS. So you’ve got to be careful in checking and setting those parameters.

One nice thing is that the structure of GeoJSON works incredibly well in typescript. It has discriminated unions built in so you can walk entire geodatasets in a pretty comfortable way.

papercrane•about 14 hours ago

> It works great as long as you squint just a bit, ignoring that it generally calls for long,lat and is designed with the assumption of a world CRS.

I thought the spec allowed you to specify the CRS, but I just checked the RFC and they removed that from the 2016 specification and WGS84 is specified. It does allow for alternative CRS with prior arrangement, but like you said that does require a lot of care.

CornCobs•about 2 hours ago

Yes, they deprecated the CRS field and the current state of geojson handling libraries is pretty messy as a result since geojson does not have versioning!

If you have old geojson in a different projection, will your library respect the crs field or will it simply misinterpret your data?

Wondering if anyone could shed light on the decision to remove it as a standard when projection seems to be a critical part of GIS.

Waterluvian•about 2 hours ago

Coordinate systems and projections are one of those deeply complicated truths that makes such a headache in GIS. I still shudder at all the pain in school and at previous jobs dealing with inconsistent datasets.

It seems like they decided to just opt out of trying (see the yellow box in section 4): https://stevage.github.io/geojson-spec/#section-4

I think they should have completely backed off from touching on projections and datums in the format altogether. Ie. Something like, “coordinates are 2 or 3 tuples where the values in order correspond with easting/long and northing/lat and elevation/altitude. See metadata for agreed upon units and CRS/projection semantics. It is strongly encouraged to standardize on WGS84 when encoding data with an earth-resolvable datum.”

Because GeoJSON otherwise works fine for indoor spaces, video game spaces, fictional lands, other celestial bodies, etc. You just have to educate on the idea that there’s more to data compatibility than it being GeoJSON.

drewda•about 9 hours ago

Yup, technically speaking if the coordinates aren't in WGS84, it isn't GeoJSON

matt-p•about 11 hours ago

OK, I had not considered just using GeoJSON for my flatland CRS (indoor routing). Quite obvious in hindsight, thank you.

sam_lowry_•about 14 hours ago

> tens of thousands of warehouse robots

Sounds like Amazon

Waterluvian•about 14 hours ago

Definitely not Amazon. Yuck.

DarkNova6•about 14 hours ago

I’ve had nothing but problems using GeoJson. The specification has limitations everywhere and doesn’t even support z + m values at the same time.

But thankfully there is also the SQLite backed GeoPackage, which is not only more flexible but also much smaller. It takes some extra steps to get testing teams working due to it’s binary nature, but other than that it is the best format in geospatial data analysis.

Long live SQLite!

jackconsidine•about 15 hours ago

GeoJSON is super useful. At Getcho (delivery, logistics) we use zip code GeoJSON encodings to draw polygons on zone maps and quickly generate rates. This has been a persistently annoying thing to do until we discovered this format. If you're curious, someone made a repo with all the 2010 census zips a while back [0].

[0] https://github.com/OpenDataDE/State-zip-code-GeoJSON/blob/ma... although you can generate newer versions from the last census.

korkoros•about 15 hours ago

About 25% of ZIP codes don't have a corresponding Census Bureau ZCTA, for example 10118. Do you end up needing special handling for those cases? Or has it not yet come up in practice?

jackconsidine•about 14 hours ago

Excellent question it certainly does come up. Practically speaking the more populous zip codes are all accounted for and that’s where the vast majority of deliveries go to. For example I took the census zip code data 150 miles (crow flies) outside Philly and found virtually 100% coverage.

For missing ones you have to fall back to distance based estimates and in my business that means you’re quote may be off and you’re exposed

ryandrake•about 12 hours ago

No shade whatsoever at you or your business: I'll say upfront that you certainly made the right practical decision for the goal of running a business.

That said, this is a textbook example of what I have always found so infuriating, personally, about working on commercial software, and one of the many reasons I ultimately moved into a non-software-writing role. The (very sensible and practical) shortcuts and tradeoffs that are commonly made due to time and cost constraints. The attitude of "well the vast majority of our use cases work, so we're done." I've always thought edge cases must be addressed. Something in my brain hurts when I knowingly release something where only 99% of cases work.

I can imagine this is probably the same thing some artists feel when they are commissioned to produce (in their view rushed, flawed, or incomplete) artwork for business purposes.

I only write software at home, as a hobby now, and this gives me the outlet to follow my heart around edge cases!

cr125rider•about 12 hours ago

Made by Sean Gillies and a few others. Back when mapbox was doing all sorts of great open source stuff. Legends

https://github.com/sgillies

nobleach•about 14 hours ago

We used this extensively when I worked in this space (2010 - 2014). My favorite addition was using https://github.com/topojson/topojson to add arcs. That cut down on quite a bit of points to represent curves.

jtbaker•about 13 hours ago

Dang, fun memories of when I was first getting in to geo/data stuff and doing a lot of web mapping stuff with D3, Leaflet and friends. Seems as tools like Vector tiles/PMTiles have supplanted topojson for a lot of visualization oriented use cases.

nobleach•about 12 hours ago

I'm gonna have to dive into a rabbit-hole! I was working on an ESRI Shapefile to GeoJson converter back in those days. But D3 and Leaflet were such cool tech! MapBox too. Linking SagaGIS with PostGIS to do pre/post wildfire analysis was my jam.

ragebol•about 14 hours ago

Have been using GeoJSON, very handy and human-readable, but we recently switched to GeoPackage files, as it allows for different layers, each with a different schema for additional data.

GeoPackages also allow to set a proper CRS, which is not as easy in GeoJSON IIRC.

Getting your CRSes wrong is fun...

michaeljhg•about 15 hours ago

Also https://postgis.net/

thibautg•about 14 hours ago

And with PostgREST [0], you can automatically convert any PostGIS table (with geometry or geography column) to GeoJSON by using an "Accept: application/geo+json" header in the request.

[0] https://docs.postgrest.org/en/v14/how-tos/working-with-postg...

pramsey•about 11 hours ago

At the SQL level, the ST_AsGeoJSON(record) variant will convert a tuple that includes a geometry and any combination of other columns into a GeoJSON output.

steve-chavez•about 9 hours ago

Many thanks for your work pramsey. We use that exact function [1], do you have any plans for a similar function for TopoJSON? One that also has a record parameter? [2].

[1]: https://github.com/PostgREST/postgrest/blob/f1d0e8ea2266077d...

[2]: PostGIS has https://postgis.net/docs/AsTopoJSON.html but it doesn't take a record.

Zambyte•about 14 hours ago

Also https://github.com/timescale/timescaledb

I've found it very useful for storing geospatial data over time.

pramsey•about 11 hours ago

MobilityDB might also be of interest, for people handling trajectories.

cogman10•about 11 hours ago

Interesting but, IMO, probably one of the worst uses of JSON. The data you would want to consume is already not "human readable" so it instead introduces a lot of bloat for really no benefit.

If you have a non-insignificant amount of data points to track this is going to eat just a ton of memory while also being pretty slow to encode/decode.

Imagine, for example, if we encoded this as a binary. First 2 bytes for the feature type, second 2 bytes for the geometry type, 3 bytes for a fixed point x, 3 bytes for a fixed point y, and you could optionally provide the properties as a json blob in a trailing string. That's 10 bytes for all the coordinate stuff. Less bytes than what currently stores the `"type": "Feature"` string.

morganherlocker•about 7 hours ago

> If you have a non-insignificant amount of data points to track this is going to eat just a ton of memory while also being pretty slow to encode/decode.

This is a fair critique, however, for any large GeoJSON, the coordinate arrays will dominate the size. I think it's also safe to assume this data will be gzipped at rest and over the wire, which will eliminate most of the "header" metadata size you mention. As you point out, it would be much more efficient to have a binary format, and there are good examples like these, that are ~2-3x smaller in benchmarks:

https://flatgeobuf.org/ https://github.com/mapbox/geobuf

That said, I think GeoJSON should be compared against other human readable formats like KML, which has a lot of wasted space as well, while being more difficult to read/write.

doginasuit•about 11 hours ago

Do you mean geocoordinates when you say not human readable? Those are obviously at the heart of geospatial information but there is quite a bit more to the spec that does benefit from being human readable, and I'd include longitude/latitude among them. There are also solutions like cbor which allow them to be transferred and decoded/encoded from binary. For performance critical data you can also use something like protobuf, but it would be a huge pain to handle everything that way. Json is a great choice as a general spec.

layer8•about 10 hours ago

Somewhat related: Falsehoods Programmers Believe About Map Coordinates: https://news.ycombinator.com/item?id=24659039

phillc73•about 13 hours ago

GeoJSON is not just for geographical features! Shapes of any kind work just as well.

QuPath[1], a tool for digital pathology whole slide image analysis, can export annotations in GeoJSON format (and import too I suppose).[2] This makes it really very easy to make annotations transportable between tooling.

[1] https://qupath.github.io/

[2] https://github.com/qupath/qupath-docs/blob/main/docs/advance...

tomtomtom777•about 10 hours ago

The spec states:

> The coordinate reference system for all GeoJSON coordinates is a geographic coordinate reference system, using the World Geodetic System 1984 (WGS 84) [WGS84] datum, with longitude and latitude units of decimal degrees.

So that seems to be a misuse of the format. Using a geojson library for this may get you into trouble with ranges or antimeridian cutting.

sam_lowry_•about 15 hours ago

Dunno whose website this is, but the format itself is great, and it allows for a relatively compact and relatively human-readable presentation.

A few weeks ago I (vibe)coded mxmap.be and if not for the ubiquity of geojson, it would have taken me significantly more effort.

rippeltippel•about 15 hours ago

Nice work with mxmap. It's a very good way to appreciate to what extent EU depends on US - email providers being just one of several dimensions.

flir•about 13 hours ago

What a great project.

I vibe coded something similar (different data source) with codex that went something like SQLite->GeoJSON->Leaflet and it was a dream - almost no corrections necessary. It even went off and found a really nice colour scheme for me.

biosboiii•about 13 hours ago

I love GeoJSON :) You can bring any Geo/GIS from 0 to visualization by just parsing it into GeoJSON.

geojson.io is a great editor/viewer by Mapbox. Also https://kepler.gl/demo is great for additional filtering, visualizations like heatmaps, arcs etc.

A extension to GeoJSON that works with JSONL-like semantics would be great for huge files, but this could also be solved by tiling.

kitd•about 10 hours ago

There's a map facility not linked here that allows you to build GeoJSON graphically:

https://geojson.io/#map=12.42/51.50593/-0.13003

CamouflagedKiwi•about 14 hours ago

This is nice. I haven't worked with GIS data for ages but I really like the idea of a well-understood plain text container for it. Much nicer than wrangling with binary formats like shapefiles, especially when something goes wrong and you're not sure if it's your code (well more precisely your usage of whatever library you've got for it) or the data.

jasoncartwright•about 10 hours ago

Have been using geojson for a while and was frustrated at lack of simple web tools play with it in browser. So I built https://geojson.page

larodi•about 10 hours ago

One should be aware that Google, even though JSON is JSON, would sometimes use its own binary encoding for the content of polylines and generally large sets.

dnnddidiej•about 13 hours ago

Looks like what any sensible dev would come up with if asked to "return this geo data as json". I like simple!

mtucker502•about 14 hours ago

The properties key is plural but contains a dictionary. Does the schema allow for this to be a list?

kijin•about 14 hours ago

Nope, properties must be an object (dictionary or null). Which means each property can only appear once.

The spec doesn't say what type the value of a property can be, though. Examples in the RFC show strings, floats, and a nested object. So you could probably put a list in there as well if you want to store multiple values under the same key, provided that your decoder knows what to do with such values. (GeoJSON is often converted to and from WKB/WKT, and unorthodox values may be lost in the conversion.)

tosh•about 15 hours ago

vega-lite supports rendering of GeoJSON via 'geoshape'

https://vega.github.io/vega-lite/docs/geoshape.html

trgn•about 13 hours ago

nice and simple, great. but because it's json, most parsers are horribly inefficient, which is tough, because a lot of geodata is massive.

jeffbee•about 12 hours ago

JSON parsing is probably one of the most thoroughly optimized subsystems in the whole industry at this point. Obviously there are ways to encode the same data that are easier to parse (e.g. instead of absolute floating point coordinates, use integer deltas along a path in some reasonable CRS) but because this inefficient representation is so common and so long-standing the parsing is faster than people think.

trgn•about 12 hours ago

the default parsers all load the entire thing in mem, which is not good.

so you need a stream-based parser, which nobody does an effort to write/use for json. especially since geojson is a web format, and people just default to json.parse, which is blocking. and even then, even if you did use the custom one, it likely won't be a geojson-tailored one, so because key-order isn't guaranteed, any parser for geo-json will need to do some acrobatics to finding the reference-system, dealing with arbitrarily nested geometries etc..

it's a good format for what it is, but it's not a great geo-format. a geo format needs to be easily scannable and, even better, have a geometry index to be able to seek quickly.

kbolino•about 7 hours ago

For whatever it's worth, you don't have to write anything special to handle the reference system, because the final, published version of RFC 7946 only allows WGS84 anyway.

vortegne•about 14 hours ago

Recently I got into cartography software for a bit and the horrors of the data formats in this industry are real. Feels like everyone under the sun has their own.

All that said, GeoJSON was a great change of pace, I enjoyed using it. While I'm no professional and have no idea what the professional needs are, it was very good for my hobbyist needs.

cyberax•about 8 hours ago

To add a bit of negative here: the format is incredibly inefficient in JS, because each point gets expanded into a full-blown JavaScript object.

You can save a lot of RAM by using an array of interleaved coordinates. For an additional bonus, you can also compress rings by storing the ring offsets inside a larger array.

Demiurge•about 13 hours ago

Also, JSON! Wow.