Comparing
less-magic
(3057cc8
) withmain
(640e8a1
)0%
- Improvements
- 0
- Regressions
- 0
- Untouched
- 7
- New
- 0
- Dropped
- 0
- Ignored
- 0
Benchmarks
Passed
test_jit[add]
python/tests/test_array_api.py::test_jit[add]0%
335 ms
333.5 ms
test_jit[tuple]
python/tests/test_array_api.py::test_jit[tuple]0%
580.4 ms
578.8 ms
test_run_lda[array_api-optimized-numba]
python/tests/test_array_api.py::test_run_lda[array_api-optimized-numba]0%
239.6 µs
239.5 µs
test_jit[lda]
python/tests/test_array_api.py::test_jit[lda]0%
7.6 s
7.6 s
test_run_lda[array_api-optimized]
python/tests/test_array_api.py::test_run_lda[array_api-optimized]0%
874.3 µs
875 µs
test_run_lda[array_api]
python/tests/test_array_api.py::test_run_lda[array_api]0%
4.6 ms
4.6 ms
test_run_lda[base]
python/tests/test_array_api.py::test_run_lda[base]0%
2.2 ms
2.2 ms
Commits
Click on a commit to change the comparison rangeBase
main
640e8a1
0%
Remove automatic extraction in builtin evaluation
In https://github.com/egraphs-good/egglog-python/pull/265 (released as
9.0.0) the ability to automatically extract a builtin using a global
egraph context was added. This PR removes that feature, requiring all
builtins to be in a normalized form.
I realized that for https://github.com/egraphs-good/egglog-python/issues/241
we want facts to compare structural equality when converting to a boolean,
instead of using the e-graph. Looking at that previous PR, it seems like
a mistake to add this implicit context, making things more confusing
and opaque with minimal UX improvements.
7f7a52b
5 months ago
by saulshanabrook0%
Fix bug around improperly upcasting during ==
3057cc8
5 months ago
by saulshanabrook