awklet is an implementation of the core awk language in pure Rust. It was primarily designed to be an exercise for myself, and as such it uses no pre-made parsing or lexing libraries; the entirety of the interpreter is implemented from scratch. That being said, it may be of use to others, and exposes individual interfaces for lexing, parsing and evaluation.
awklet is thoroughly covered by unit tests and a handful of integration tests.
To the right: an auklet, one species of auk. Source: Wikipedia.
awklet is split into three parts: the lexer, parser and evaluator.
Maps an input string to a stream of tokens. It is entirely context free, and rather naive in its structure. This is largely appropriate for the awk language, with minor exceptions. Notably, there are existing shortcomings in its ability to discern division from regex literals. This functionality is implemented as a rather ugly context-dependent hack in the gawk grammar, and I deemed it not sufficiently important to sacrifice simplicity in awklet.
Ingests a token stream to construct an AST. The bulk of this component is a standard recursive-descent parser, however groups of tokens in expression position are extracted together and evaluated by a separate operator precedence parser.
The overarching recursive parser is rooted at src/parser/mod.rs, while the operator precedence
parser lives in src/parser/expression_ops.rs.
Owns a provided AST, and exposes interfaces for feeding records into that AST's defined rules.
Structurally, the evaluator manages record-level operations and selects actions to execute, while
it defers to the internal ExecutionEngine for executing individual statements and managing the
global closure. Unlike the lexer and top-level parser components, the evaluator is stateful and
internally-mutable. The current record, along with active variables, are members of the closure.
- Default,
BEGIN,END, expression, and range patterns.- Range patters are still to-do at the evaluator level. Regex are not currently implemented.
print- Variables and all three core types (numeric, string, and "numeric string")
FS,OFS,RS,ORS,NF,NR- Correct behavior of changing fields, and their interop with
NFandORS - Most of the language's operators, minus gawk or mawk extensions
- There is currently no frontend CLI.
- Most of the standard library functions (all functions aside from
length)- Parser support is present, and implementing these functions would be straightforward.
- Live user input, e.g.
getline - User-defined functions
- Control flow constructs, e.g.
if,for - Piping or IO redirection
- Arrays