8c73e4bb29
Eliminate the duplication between `constructor_declaration` and
...
`bar_constructor_declaration`. The trick is to parameterize
`constructor_declaration` with an `opening` which can be
instantiated either with `epsilon` or with `BAR`. I am quite
proud of this one!
2018-11-20 12:28:18 +01:00
a4419a8f87
New entry `make interpret-menhir`, to help understand and debug the parser.
2018-11-20 12:28:18 +01:00
fe0b3f3dfb
In `Makefile.menhir`, request the production of `parser.automaton`.
2018-11-20 12:28:18 +01:00
a6000fbf48
Inline `type_variable` and `anonymous_type_variable` back into
...
`optional_type_variable`.
2018-11-20 12:28:16 +01:00
0d328a7c83
Changed the definition of `formal_class_parameters` to allow the use
...
of an anonymous type variable as a formal parameter to a class.
This enlarges the language, and can be considered a bug fix, as
the printer (`ocamlc -i`) could actually produce this form.
Added a test case.
2018-11-20 11:52:47 +01:00
a8edab5857
Introduce `anonymous_type_variable`.
...
Simplify the definition of `optional_type_variable`.
2018-11-20 11:52:47 +01:00
8db0892294
Mark `type_variable` as `%inline`.
2018-11-20 11:52:47 +01:00
ea77a2739f
Inline `optional_type_parameter_list` at its single use site.
2018-11-20 11:52:47 +01:00
f25bc5215d
Renamed `class_type_parameters` to `formal_class_parameters`.
...
Introduced `class_parameters(parameter)` so as to improve sharing
between `formal_class_parameters` and `actual_class_parameters`.
2018-11-20 11:52:47 +01:00
02ac26414c
Inline `type_parameter_list` at its single use site in `class_type_parameters`.
2018-11-20 11:52:47 +01:00
6334a7e2f7
Relax the definition of `row_field` by replacing `atomic_type` with
...
`core_type`. This allows, e.g., removing the parentheses in the
variant type `[ | (foo [@hello]) ]`. This change is somewhat pointless,
but requiring parentheses here was pointless, too.
2018-11-20 11:52:47 +01:00
e55707fd13
Remove an inaccurate comment. Add a CONFLICTS file with explanations.
2018-11-20 11:52:47 +01:00
e942542311
Comments.
2018-11-20 11:52:47 +01:00
89e70c1b22
Simplified definition of `post_item_attributes`.
2018-11-20 11:52:47 +01:00
cd36487fa0
Simplified definition of `attributes`.
2018-11-20 11:52:47 +01:00
60f67fb500
Introduce `alias_type` as an internal name for `core_type_no_attr`,
...
an external concept (a core type without attributes).
2018-11-20 11:52:47 +01:00
18fc9477ef
Change `function_type` to `core_type_no_attr` in the definition
...
of `core_type_no_attr`. This means that `int as 'a as 'b` is now
accepted as a valid core type, whereas it was previously rejected.
2018-11-20 11:52:47 +01:00
1152ef1c7f
New symbol `arg_label`, for improved sharing and clarity.
2018-11-20 11:52:47 +01:00
737e79d8be
Change the domain of function types from `function_type` to `tuple_type`.
...
This reduces the number of conflicts. A `%prec` annotation must be moved.
2018-11-20 11:52:47 +01:00
fc5619eb60
Add a new test, which exercises the resolution of a shift/reduce
...
conflict on the symbol MINUSGREATER.
2018-11-20 11:52:47 +01:00
be2d62e3ee
`optlabel` does not need to be marked `%inline`.
2018-11-20 11:52:47 +01:00
8c083b3e60
Improve sharing by introducing `optlabel`.
2018-11-20 11:52:47 +01:00
9ee30e8e9e
Rename `core_type2` to `function_type`. Add a comment.
2018-11-20 11:52:47 +01:00
886c952e44
Inline away `core_type2_`.
2018-11-20 11:52:47 +01:00
fe07c6d3ba
Define the auxiliary symbol `extra_rhs`. Inline away `extra_core_type2`.
2018-11-20 11:52:47 +01:00
74307436b1
Rename `atomic_type_or_tuple` to `tuple_type`. Add a comment.
2018-11-20 11:52:47 +01:00
8abc20e64e
Changed the definition of `atomic_type_or_tuple`
...
so as to use `separated_nontrivial_llist`.
As a result, `atomic_type_star_list` is used only in one place
and can be manually inlined into `constructor_arguments`.
2018-11-20 11:52:47 +01:00
814ca1a636
Remove `core_type_comma_list`, which was used in only one place
...
and therefore can be manually inlined.
2018-11-20 11:52:47 +01:00
78d079db3e
Improve sharing by introducing `actual_class_parameters`.
2018-11-20 11:52:47 +01:00
dc0029c8fe
Reformulate the definition of `actual_type_parameters`
...
to use `separated_nontrivial_llist`. Although this makes
the automaton slightly larger, it is a higher-level
definition. Added a comment.
2018-11-20 11:52:47 +01:00
5de77f8dc2
Rename `core_type_list` to `atomic_type_star_list`.
2018-11-20 11:52:47 +01:00
8de2d64250
Merged `simple_core_type`and `simple_core_type2`,
...
and renamed the resulting symbol `atomic_type`.
This is made possible by using a list of *two* or more
core types in the definition of `actual_type_parameters`.
As a result, `inline_core_type_comma_list` is no longer
useful and disappears.
Some directives of the form `%prec below_HASH` must be
moved as a result of this change.
2018-11-20 11:52:47 +01:00
8abe28dbfd
Add a test in `tests/parsing` which exercises the resolution of
...
(some of) the shift/reduce conflicts involving the token HASH.
2018-11-20 11:52:47 +01:00
cc85be6e1b
Renamed `simple_core_type2` to `atomic_type`. Added a comment.
2018-11-20 11:52:47 +01:00
88c29fb401
Turn `simple_core_type2_` into an anonymous symbol.
...
This is equivalent to marking it `%inline`, but is more readable.
2018-11-20 11:52:47 +01:00
96883a1af1
Introduce the nonterminal symbol `actual_type_parameters` so as
...
to improve sharing in the definition of `simple_core_type2_`.
2018-11-20 11:52:47 +01:00
8e99cefb28
Update the definitions of `core_type_comma_list` and
...
`inline_core_type_comma_list` so as to remove the need
for explicit calls to `List.rev`.
2018-11-20 11:52:47 +01:00
b0770b6d5a
Add a definition of `inline_separated_nonempty_llist`.
2018-11-20 11:52:47 +01:00
de8df5dbb0
Exclude `boot/menhir/parser.{ml,mli}` from the list of source files
...
used by `make build-all-asts` and related commands.
Indeed, the idea is to check that the ASTs do not change when the parser is
modified -- the parser's source itself, if included, would always produce a
false positive.
2018-11-20 11:52:47 +01:00
83430b0434
Parser: introduce `virtual_with_mutable_flag` and `virtual_with_private_flag`
...
in order to eliminate a redundancy between productions in two places.
This removes two TODO markers.
2018-11-20 11:52:47 +01:00
de4915e715
Parser: remove several uses of `syntax_error()` by explicitly using
...
`no_override_flag` instead of `override_flag` when this flag is disallowed.
A conflict is avoided by decorating `override_flag` with `%inline`.
2018-11-20 11:52:47 +01:00
e987491bcb
Make the grammar more precise by explicitly disallowing `NONREC`
...
inside `str_type_extension` and `sig_type_extension`. This removes
the need for post-hoc checks, and isolates two calls to `not_expecting`.
2018-11-20 11:52:47 +01:00
87883b6f92
Parser: parameterize `let_binding` and `let_bindings` with `EXT`.
...
`EXT` can be instantiated with either `ext` of `no_ext`.
This allows getting rid of one use of `not_expecting`
in the auxiliary function `class_of_let_bindings`.
This use of `not_expecting` is now clearly isolated
in the definition of `no_ext`.
2018-11-20 11:52:47 +01:00
4db3221aad
Inline `ext_attributes` into `let_binding`.
2018-11-20 11:52:47 +01:00
de3dcf1e2e
Create [make demote-menhir] to undo the effect of [make promote-menhir].
2018-11-20 11:52:47 +01:00
f27c1de1df
Parser: A simpler definition of `ext_attributes`.
...
(Not sure why a complicated definition was used.)
2018-11-20 11:52:47 +01:00
78590941b9
Convert ephe-c-api tests (GPR#676) to ocamltest
2018-11-18 22:46:28 +00:00
de99636e8a
Merge pull request #2152 from gasche/typedecl-split-fixpoints
...
typing/typedecl: split and factorize fixpoint computations for variance and immediacy
2018-11-18 18:55:13 +01:00
7b29c553fd
typedecl: refactor variance requirements to present a unified pipable interface
2018-11-18 15:59:34 +01:00
1128d66c8a
typedecl: remove the location from variance requirements
...
The location can always be reconstructed from `decl.type_loc`. Note
however that variance-computation functions, in particular
`compute_variance_extension` and its callee (`compute_variance_gadt`,
etc.) need to be passed both the requirement and the location, despite
the fact that they also take a `decl` parameter: there are other
use-cases where the location is not the location of the type
declaration -- when processing an extension to an existing type
declaration.
2018-11-18 15:42:14 +01:00
François Pottier
David Allsopp
Armaël Guéneau
Gabriel Scherer