It is not valid syntax for a record extension to be anything except a
variable, so I changed the...

Again, it is the same as 0.18 this way, so it should not block 0.19

To get things done, it makes sense to keep 0.18 quality as the bar I
must achieve. That means I ...

The idea is that type inference should produce AST.Canonical.Type if it
succeeds. It can strip o...

Thanks to https://github.com/elm-lang/error-message-catalog/issues/193
for the tip on this!

Going to do things differently in different phases

Much more specific now. And they stress that it’s necessary to change
mindset, and not just grow...

Get indentation and coloring right again.

For example, the link to elm-lang.org is updated to be shorter and
nicer.

Allow docs on ports. It should be fine. Publishing modules with ports
will be ruled out elsewher...

Mutually recursive definitions may include tail-call definitions, so I
gotta handle that case too!

It was dropping all of the content! New versions keeps it. Definitely
takes up more bits, but it...

This was caught in some issue.

Get Debug.log working

The only important change is in Opt.Case, but the changes about
destructuring lists and tuples i...

It may be used by kernel code, so it makes sense to generate and let
the JS minifiers get rid of...

Now it handles unboxing and enums!

This means I can modify the implementation within the Elm.Name module
without messing with every...

This is prep work that will allow experimenting with alternate
implementations.

This matches 0.18 behavior

I do not really like this, but I need the dependency and field
information from generating the J...

Payload type is needed to generate encoders and decoders, whereas the
overall type is needed for...

It is now defined in the String module, so it needed to be adjusted a
bit.

The problem was that in the old version, the path to an Alias was
different than a path to other...

There are two main things here:

1. Gotta use the short name. No Elm.Kernel. prefix.
2. Register...

In the old one, having two imports triggered the “this is ambiguous
error” so I changed it to ma...

The info produced in Left was the same for everything. I.e. it was
useless and can be made more ...

This has two major benefits:

1. I can keep track of field order in objects.
2. I can give nice ...

Also remove ~ from the valid symbols. Not needed and generally too
close to - in small sizes.

It used to put $ between all variable names, but if there was only one,
it did not create a new ...