When removing fields, rename new ones

[perrin4869]

If the array of fields have different shapes, removing a field will result in new fields with corrupt state. According to

final-form-arrays/src/moveFieldState.js

Line 15 in 7b37726

// when that field gets registered, with its own change/blur/focus callbacks

, the new field will get registered, and the correct functions will be added, but unfortunately, final form never does, since the following condition will fail: https://github.com/final-form/final-form/blob/4476c4539c9a9514a1ecbfdade974bd0aafd3b29/src/FinalForm.js#L775

I tried asserting that those 3 functions always exist, by adding the following lines to final-form's registerField:

// in case a mutator like final-form-arrays moves a field state without registering
state.fields[name].blur = state.fields[name].blur || (() => api.blur(name))
state.fields[name].change = state.fields[name].change || (value => api.change(name, value))
state.fields[name].focus = state.fields[name].focus || (() => api.focus(name))

And this fixed the issue too. But I thought this approach is a little bit better

[codecov]

Codecov Report

Merging #39 into master will not change coverage.
The diff coverage is 100%.

@@          Coverage Diff          @@
##           master    #39   +/-   ##
=====================================
  Coverage     100%   100%           
=====================================
  Files          12     12           
  Lines         139    141    +2     
  Branches       29     30    +1     
=====================================
+ Hits          139    141    +2

Impacted Files	Coverage Δ
src/remove.js	100% <100%> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 7b37726...572f5fe. Read the comment docs.

[perrin4869]

I think that maybe, this change should be moved inside of moveFieldState, and remove the possibility of a null change/blur/focus state altogether. But that is a slightly bigger change and breaks more tests

[erikras]

Published in v3.0.2.