Create better examples

[letmaik]

Example coverages should be more meaningful/realistic, but still simple/small.

[letmaik]

@jonblower Any idea? I noticed this as a problem repeatedly while showing it to people. The big grid cells look weird for example.

[jonblower]

Barbara is working on a sample flooding dataset that could be useful. And we have some coastal-ocean data, which isn't too big but still looks nice. We'd have to check the licence conditions. Any idea on suitable max size?

[letmaik]

The maximum size should be such that it comfortably fits into a standalone covjson document that can be displayed in the editor. So not super big. Up to 300-400 domain objects are probably fine, for a regular axis Grid. Other domain types should be way smaller where we can't use regular axes.

On 18/07/2016 17:26, Jon Blower wrote:

Barbara is working on a sample flooding dataset that could be useful. And we have some coastal-ocean data, which isn't too big but still looks nice. We'd have to check the licence conditions. Any idea on suitable max size?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/covjson/playground/issues/1#issuecomment-233362946, or mute the thread https://github.com/notifications/unsubscribe-auth/AAgaLO_fo4eV5sfDwKe38xmFX_wvRRv4ks5qW5tAgaJpZM4HVYZn.

[jonblower]

So I guess that means a 20x20 grid, roughly speaking (sqrt 400)? But for a regular axis grid, you could have as big a grid as you want and it would be the same size in the encoding? Have I misunderstood?

[letmaik]

Sure, but the range values have to be in the document as well. I'd rather not have URL ranges for the examples, at least not for the majority.

On 18/07/2016 17:55, Jon Blower wrote:

So I guess that means a 20x20 grid, roughly speaking (sqrt 400)? But for a regular axis grid, you could have as big a grid as you want and it would be the same size in the encoding? Have I misunderstood?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/covjson/playground/issues/1#issuecomment-233371988, or mute the thread https://github.com/notifications/unsubscribe-auth/AAgaLDFxkv4eDtBZmGlIysZYwYruq1GUks5qW6HogaJpZM4HVYZn.

[jonblower]

Right. Hard to make it look "real" with such a small dataset but you could pick a pattern, like concentric circles or an "egg box" pattern. Or take something simpler like a gradient fill and perturb it a little.

If you only want a dataset that size then the grid cells are going to appear pretty big on screen. If you want to reduce their geographical extent (to work at local scale rather than continental), you could do something like air quality within a city (but that's likely to be quite noisy), or maybe try to get data on an urban heat island.

[letmaik]

I think 20x20 is enough for a temperature grid of one country. If we pick UK, then you can also show missing values (where there's ocean). Is there any free/public weather dataset that we could downsample? Could be average air temperature of a specific day.

On 18/07/2016 23:23, Jon Blower wrote:

Right. Hard to make it look "real" with such a small dataset but you could pick a pattern, like concentric circles or an "egg box" pattern. Or take something simpler like a gradient fill and perturb it a little.

If you only want a dataset that size then the grid cells are going to appear pretty big on screen. If you want to reduce their geographical extent (to work at local scale rather than continental), you could do something like air quality within a city (but that's likely to be quite noisy), or maybe try to get data on an urban heat island.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/covjson/playground/issues/1#issuecomment-233463320, or mute the thread https://github.com/notifications/unsubscribe-auth/AAgaLCcvB9ISo_KuTcgw3h7cteV9JCIwks5qW-7BgaJpZM4HVYZn.