15 releases (breaking)
0.11.2 | Aug 30, 2021 |
---|---|
0.11.1 | Jun 9, 2021 |
0.11.0 | May 12, 2021 |
0.7.0 | Mar 24, 2021 |
#904 in Science
16,499 downloads per month
Used in prover-lab
7MB
8K
SLoC
z3tracer
This crate provides an experimental parser and analyzer for detailed log files produced by Z3.
To obtain a file z3.log
, pass the options trace=true
and proof=true
on the
command line of Z3 (for instance, z3 trace=true proof=true file.smt2
). For large problems, you may
choose to limit the size of the log file by interrupting z3
with ^C
or by setting
a shorter timeout. Passing only trace=true
is also possible but it will prevent any
dependency analysis (see below).
While parsing the logs, a "model" of the logs is constructed to record most operations, as
well as the syntactic terms under each #NUM
notation:
use z3tracer::{Model, syntax::{Ident, Term}};
let mut model = Model::default();
let input = br#"
[mk-app] #0 a
[mk-app] #1 + #0 #0
[eof]
"#;
model.process(None, &input[1..])?;
assert_eq!(model.terms().len(), 2);
assert!(matches!(model.term(&Ident::from_str("#1")?)?, Term::App { .. }));
assert_eq!(model.id_to_sexp(&BTreeMap::new(), &Ident::from_str("#1").unwrap()).unwrap(), "(+ a a)");
Besides, the library will attempt to reconstruct the following information:
-
Quantifier Instantiations (QIs), that is, which quantified formulas were instantiated by Z3 and why;
-
The successive backtracking levels during SMT solving;
-
SAT/SMT conflicts and their causal dependencies in terms of QIs;
-
Causal dependencies between QIs.
A tool z3tracer
based on the library is provided to process a log file z3.log
from the
command line and generate charts.
See also in the repository for additional examples using Jupyter notebooks.
Currently, this library only supports Z3 v4.8.9.
More information about Z3 tracing logs can be found in the documentation of the project Axiom Profiler.
Contributing
See the CONTRIBUTING file for how to help out.
License
This project is available under the terms of either the Apache 2.0 license or the MIT license.
Dependencies
~6–10MB
~152K SLoC