Allow invalid styles and show errors

[orangemug]

Editing a Maputnik style can be a pain right now. In many places entering an invalid value will cause the input to get in a funny state or just refuse to let you complete the action. For example changing a color input with the value rgba(255, 0, 0, 1) to rgb(255, 0, 0) and you'll see the issue.

Also editing a style via the JSON editor is next to impossible at the moment for similar reasons.

I think we should allow errors to be present in the style, and indicate what layers have errors.

Here's a prototype https://696-84182601-gh.circle-artifacts.com/0/artifacts/build/index.html#0.26/0/0 (working branch master...orangemug:feature/allow-invalid-values)

Below is an example where I've modified a layer type to be invalid, backgroun instead of background

Below is an example where I am part way through modifying a color from hsl to hsla and am still missing the forth argument.

Any thoughts or feedback welcome and encouraged 😄

[gregwolanski]

This issue appears in the answers to question “What annoys you about Maputnik?” in the survey (original spelling):

• sometimes ctl+v doesnt work if theres is mismatch,it should give error by underline or something

• Black Screen with often no error messge once you make a mistake editing the local style file.

• Bugs to refresh style (I need to play with JSON to fix issues)

• When I want to set “Offset” to “1.7” and I type “1.6” by mistake, click “Backspace” to change “6” to “7”, I get “17” instead of “1.7”, because Maputnik changes “1.” to “1” after the “Backspace”.

[pathmapper]

I'm all for it. This would allow also to change the source for an existing layer which is not possible currently.

Would it be possible in the JSON editor to mark additionally the line(s) which is/are invalid (e.g. make the line number red)?

[orangemug]

Would it be possible in the JSON editor to mark additionally the line(s) which is/are invalid (e.g. make the line number red)?

Hmmm I'm not sure, but it's a good idea I'll take a look

[pathmapper]

There is already something similar to indicate parse errors:

[sk1p]

Any thoughts or feedback welcome and encouraged

I like it! Maybe in addition to the layer-wide error, also highlight the field that is currently invalid.

[pathmapper]

There are two regressions in master since v1.5.0 which are likely related: #484, #488

[orangemug]

Closed by PR #620