A little simpler this time.

Fingers crossed.

This should be ok now that I understand how `prepublish` is supposed to
work.

Too strict.

Just testing that it's possible to manipulate row state.

This worked in the first column by accident as the queries returned zero
always. Now it seeks th...

It's better just to return source and target instead of indices.

It's the same idea as for `components`. Not fully implemented, though
(missing row/cell level ho...

Now it won't lose possible passed classNames.

I need to rethink how to handle that. Given it's a monorepo, consuming
from Git doesn't make tha...

Not needed given npm will generate one.

Otherwise React DnD fails.

It's better to retain the old styling. Otherwise toggling columns resets
it.

It's possible a column might be missing a property so better prepare for
that.

There's no guarantee a label exists so that won't work as a key.

This probably needs to become more flexible, but this is a start.

Looks like npm needs this for some reason.

This can be extracted as a package now.

I realized this doesn't belong to the easy core. The problem is that if
it goes to the core, the...

This is better as then possible performance improvements related to
columns won't break the code...

Maybe npm picks up the readme after this.

Not needed anymore.

This is just to avoid uploading junk to npm.

It's better if only `reactabular` runs that. Otherwise it will try to
deploy the site multiple t...

Now the component accepts `query` so that it's compatible with search.
It doesn't implement `Sea...

That aggregates the dependencies well. Easier to maintain.

Now they trigger `dist:all` properly and deploy the site.

`reactabular-easy` is going to need this.

This wraps the column resizing concern. Also integrated with
reactabular-easy and got rid of the...

This fixes "All features" delete.

Now it should generate the dist bundle and publish the site when
`reactabular` main package is u...

docs - Change "row" to "onRow"

Sadly Lerna doesn't manage this. Maybe there's some hook in Lerna to
allow triggering preversion...

It's better to go with a synchronized model (Babel model). Otherwise
it's difficult to tell whic...

A little neater this way.

This can use just `props` over `transforms`.

This way `props` work for children columns as expected. I need to add a
proper test for this cas...

It's better to tie this to the column definition. That way toggling
columns works with minimal e...

This is easier to understand in isolation.

Clearer this way.

Other resolvers might need this data so better not to lose it.

Related to #170.

I found a simpler way to pull this off. A bit hacky but this works.

Cleaner this way. The output can remain `dist-modules` as I don't think
there's a strong convent...

I'm not a great fan of this, but at least `lerna run` works now.

The broken behavior works to our favor here. Now ESLint can be improved
too (can check imports a...

`lerna bootstrap` doesn't make sense with broken `prepublish`...

The versions of the packages don't have to be in sync.