catala/compiler/surface/surface.mld
2022-01-19 10:54:16 +01:00

79 lines
3.1 KiB
Plaintext

{0 Catala surface representation }
This representation is the first in the compilation chain
(see {{: index.html#architecture} Architecture}). Its purpose is to
host the output of the Catala parser, before any transformations have been made.
The module describing the abstract syntax tree is:
{!modules: Surface.Ast}
This representation can also be weaved into literate programming outputs
using the {{:literate.html} literate programming modules}.
{1 Lexing }
The lexing in the Catala compiler is done using
{{: https://github.com/ocaml-community/sedlex} sedlex}, the modern OCaml lexer
that offers full support for UTF-8. This support enables users of non-English
languages to use their favorite diacritics and symbols in their code.
While the parser of Catala is unique, three different lexers can be used to
produce the parser tokens.
{ul
{li {!module: Surface.Lexer_common} corresponds to a concise and programming-language-like
syntax for Catala. Examples of this syntax can be found in the test suite
of the compiler.}
{li {!module: Surface.Lexer_en} is the adaptation of {!module: Surface.Lexer_common}
with verbose English keywords matching legal concepts.}
{li {!module: Surface.Lexer_fr} is the adaptation of {!module: Surface.Lexer_common}
with verbose French keywords matching legal concepts.}
}
Relevant modules:
{!modules: Surface.Lexer_common Surface.Lexer_fr Surface.Lexer_en}
{1 Parsing }
The Catala compiler uses {{: http://cambium.inria.fr/~fpottier/menhir/} Menhir}
to perform its parsing.
{!module: Surface.Parser} is the main file where the parser tokens and the
grammar is declared. It is automatically translated into its parsing automata
equivalent by Menhir.
In order to provide decent syntax error messages, the Catala compiler uses the
novel error handling provided by Menhir and detailed in Section 11 of the
{{: http://cambium.inria.fr/~fpottier/menhir/manual.pdf} Menhir manual}.
A [parser.messages] source file has been manually annotated with custom
error message for every potential erroneous state of the parser, and Menhir
automatically generated the {!module: Surface.Parser_errors} module containing
the function linking the erroneous parser states to the custom error message.
To wrap it up, {!module: Surface.Parser_driver} glues all the parsing and
lexing together to perform the translation from source code to abstract syntax
tree, with meaningful error messages.
Relevant modules:
{!modules: Surface.Parser Surface.Parser_driver Surface.Parser_errors}
{1 Name resolution and translation }
The desugaring consists of translating {!module: Surface.Ast} to
{!module: Desugared.Ast} of the {{: desugared.html} desugared representation}.
The translation is implemented in
{!module: Surface.Desugaring}, but it relies on a helper module to perform the
name resolution: {!module: Surface.Name_resolution}. Indeed, in
{!module: Surface.Ast}, the variables identifiers are just [string], whereas in
{!module: Desugared.Ast} they have been turned into well-categorized types
with an unique identifier like {!type: Scopelang.Ast.ScopeName.t}.
Relevant modules:
{!modules: Surface.Name_resolution Surface.Desugaring}