Add Pp_ast module and ppxlib-pp-ast executable for pretty-printing ppxlib ASTs

[NathanReb]

This adds pretty-printing functions to our API along with a simple executable that can be used to pretty-print input .ml, .mli, expressions, patterns or core types.

It's API is similar to ppx_tools's dumpast except that it can also read binary ASTs of any version which I assume could come in handy in certain situation.

The printing style is also similar to dumpast:

• It prints expressions directly as their pexp_desc field and does so for any other such record with a _desc field
• It does not print locations, positions or attributes to avoid cluttering the output
• It prints 'a loc types directly as 'a

In the future we can add a bit of configuration there to control what gets printed or not but it is already pretty useful as it is.

I'd even consider using this with our -dparsetree driver option instead of the current sexp printer.

[NathanReb]

I still need to add a few tests before this is good to merge.

It uses a bit of private API and re-implements a few small parts of Ppxlib.Utils that aren't exposed but given the amount of code that's impacted I thought it would be simpler to do that than to try and expose those to the executable while keeping them private to the outside world.

here's an example of how let x = 2 + 2 gets printed out:

$ dune exec -- ppxlib-pp-ast test.ml
[ Pstr_value                           
    ( Nonrecursive
    , [ { pvb_pat = Ppat_var "x"
        ; pvb_expr =
            Pexp_apply
              ( Pexp_ident (Lident "+")
              , [ ( Nolabel, Pexp_constant (Pconst_integer ( "2", None)))
                ; ( Nolabel, Pexp_constant (Pconst_integer ( "2", None)))
                ]
              )
        ; pvb_attributes = __attrs
        ; pvb_loc = __loc
        }
      ]
    )
]

I'm not a Format expert yet so there might be a few bugs or quirks but I'm hoping this can be fixed iteratively rather than waiting for it to be perfect before releasing it.

Please let me know what you think!

[NathanReb]

Just added a config option to allow showing attributes as it turns out I needed this to investigate a migration bug.

[patricoferris]

This looks great! I think waiting for the migration bug fix before merging might help uncover more useful features to add (e.g. you needed the --show-attrs configuration flag).

[NathanReb]

I'll try to address the documentation related concerns and the exposure of private API parts for internal use but then I think this is good to go!

[NathanReb]

I added loc printing and relevant configuration!

[NathanReb]

I'm also wondering whether the tool should be installed as part of ppxlib's main package or a separate, ppxlib-tools one.

[NathanReb]

I improved the documentation a bit, let me know what you think @patricoferris !

[patricoferris]

The new documentation looks great!

I'm also wondering whether the tool should be installed as part of ppxlib's main package or a separate, ppxlib-tools one.

ppxlib-tools would just be to attach some executables too?

[NathanReb]

ppxlib-tools would just be to attach some executables too?

Yep, in order to keep ppxlib minimal as not every ppx user would be interested in ppxlib-pp-ast.

[NathanReb]

In addition, we could also extend the set of dependencies for such tools to things like cmdliner for instance, whereas we try to keep the set of dependencies for ppxlib to a bare minimum.

[NathanReb]

This should be the final round, I extracted ppxlib-pp-ast to a separate package and rewrote the tool itself using cmdliner.

I also added the ability to read source code directly from the command line which can be extremely convenient to quickly understand how to represent simple expressions or types for instance. This feature was part of ppx_tools' dumpast and I quickly realized it wasn't very convenient to go through a file or stdin for the simplest use cases.

I also took the time to write a bit of documentation in the manpage so hopefully the tool will be easier to use.

[patricoferris]

This looks great to me! Happy for formatting bugs to be follow up PRs if they do arise, this would already be very useful for debugging migration issues!