nouritsu / rs-pseudocode Goto Github PK

The parser used to output a single Expr, it now outputs Vec<Stmt>, the tests need to be changed accordingly.

Moving the parser tests into parser.rs as unit tests makes more sense.

This issue tracks the implementation of an Expression evaluator.

The evaluator should walk the AST and evaluate each node.

A string literal like "Hello\nWorld" contains the escaped character \n, which is being parsed as is.

Whenever the parser tries to parse an escape sequence, a character replacement can be made. For example, replacing \t with the tab character etc.

The Chumsky Example JSON Parser uses an approach that could be adapted.

A REPL or a Read Execute Print Loop would allow users to execute small snippets of pseudocode.
The interpreter will run in REPL mode if no file is passed in.

It would be simplest to write the REPL manually.

The repository has a grammar/grammar.enbf that defines grammar rules to be implemented using chumsky.

This can be used to understand ENBF syntax and this can be used to understand Pseudocode syntax.

Tests should handle edge cases like -

• chained operators
• varying operator precedence
• grouping expressions
• invalid literals
• invalid variable names
• ... and more

The Cargo inbuilt test framework is perfect for this. There is no need to section tests off into a different folder as this is a small project.

Immediate execution flag.

The flag would be used as follows -

rs-pseudocode -e "TRUE AND FALSE"
> FALSE

It would evaluate/execute the string provided.

Currently the enum contains variants like -

Unary(Operator, Box<Expr>),
Binary(Box<Expr>, Operator, Box<Expr>),

The operator enum itself can be removed and the Expression Enum can instead contain variants -

Negate(Box<Expr>),
Not(Box<Expr>),
Multiply(Box<Expr>, Box<Expr>),
Divide(Box<Expr>, Box<Expr>)

This would make the evaluation process easier since the operator need not be matched against.

In code like -

5 + NOT

NOT is a keyword, not an identifier. The ident() parser should skip keywords.

List of keywords (maybe incomplete, will be edited) -

NOT, OR, AND, DIV, MOD,
IF, ELSE, ENDIF,
FOR, NEXT, ENDFOR,
WHILE, ENDWHILE,
PROCEDURE, ENDPROCEDURE,
FUNCTION, ENDFUNCTION, RETURNS,
CASE, ENDCASE,
ARRAY,
INTEGER, REAL, CHARACTER, STRING, DATE,

There are various features in chumsky 1.0 that would allow the parser to be more readable, optimized and extendable.

Some functions need to be rewritten completely, while rewriting others would only serve to improve readability and speed.

GitHub actions integrate extremely well with Rust.
They can be used to run tests and builds on push / PR.

As the project grows, more flags would be made available. Flags like -

--debug / -d         : show debug information while interpreting
--help / -h          : show help text
--version / -v       : show current program version
--parse-only / -p    : only parse, do not interpret
--time / -t          : show time taken to parse and execute

One of the following crates should be used -

• clap
• gumdrop
• argh

This issue tracks the implementation of the rules defined in this ENBF file using chumsky.