Allow Symbolics 6 by compiling out the Symbolics extension

[tomneiser]

The FastInterpolationsSymbolicsExt weak compat pinned Symbolics = "7", which forces the entire environment to Symbolics 7 whenever Symbolics is present (weak compat is enforced even when the extension is unused). This made FastInterpolations 0.4.x unresolvable in ecosystems still on the Symbolics 6 / SymbolicUtils 3 / ModelingToolkit 9 generation (e.g. FUSE via ThermalSystemModels), breaking downstream builds.

Guard the extension body behind isdefined(SymbolicUtils, :TypeT) (the SymbolicUtils 4 API the extension relies on) so it compiles out as a no-op on the older generation, and widen the weak compat to Symbolics = "6, 7". Symbolics 7 keeps full symbolic-interpolation functionality; Symbolics 6 environments can install FastInterpolations with the extension simply inactive. Numeric interpolation is unaffected on either generation.

Verified: precompiles as a no-op under Symbolics 6.58 / SymbolicUtils 3.32 and precompiles + functions under Symbolics 7.26 / SymbolicUtils 4.

[github-actions]

FastInterpolations.jl Benchmarks

All benchmarks (50 total, click to expand)

Benchmark	Current: 4a3d7d3	Previous	Imm. Ratio	Grad. Ratio
10_nd_construct/bicubic_2d	37230.0 ns	37551.0 ns	0.991	0.981
10_nd_construct/bilinear_2d	576.1 ns	608.9 ns	0.946	0.843
10_nd_construct/tricubic_3d	351942.0 ns	355206.0 ns	0.991	0.999
10_nd_construct/trilinear_3d	1636.1 ns	1677.1 ns	0.976	0.904
11_nd_eval/bicubic_2d_batch	1438.7 ns	1436.7 ns	1.001	0.943
11_nd_eval/bicubic_2d_scalar	16.3 ns	16.3 ns	1.0	1.001
11_nd_eval/bilinear_2d_scalar	7.5 ns	7.5 ns	1.0	1.036
11_nd_eval/tricubic_3d_batch	3239.1 ns	3251.0 ns	0.996	0.992
11_nd_eval/tricubic_3d_scalar	33.3 ns	33.7 ns	0.988	0.975
11_nd_eval/trilinear_3d_scalar	13.2 ns	13.5 ns	0.978	0.984
12_cubic_eval_gridquery/range_random	4238.6 ns	4270.4 ns	0.993	0.983
12_cubic_eval_gridquery/range_sorted	4221.9 ns	4219.3 ns	1.001	0.98
12_cubic_eval_gridquery/vec_random	9613.9 ns	9367.5 ns	1.026	0.993
12_cubic_eval_gridquery/vec_sorted	3198.4 ns	3204.8 ns	0.998	0.997
13_nd_oneshot_gridquery/bicubic_2d_rand_rand	65249.8 ns	65274.3 ns	1.0	0.984
13_nd_oneshot_gridquery/bicubic_2d_sort_rand	62577.8 ns	62137.5 ns	1.007	1.002
13_nd_oneshot_gridquery/bicubic_2d_sort_sort	53984.6 ns	58071.9 ns	0.93	0.917
13_nd_oneshot_gridquery/bilinear_2d_rand_rand	16104.9 ns	19221.4 ns	0.838	0.907
13_nd_oneshot_gridquery/bilinear_2d_sort_rand	9375.9 ns	9675.1 ns	0.969	0.959
13_nd_oneshot_gridquery/bilinear_2d_sort_sort	5283.8 ns	5715.3 ns	0.924	0.975
14_series_oneshot_batch/constant_inplace_vec_k8_q1000_rand	16585.2 ns	19236.0 ns	0.862	0.869
14_series_oneshot_batch/linear_inplace_vec_k8_q1000_rand	17505.0 ns	17920.6 ns	0.977	0.931
1_cubic_oneshot/q00001	545.8 ns	540.2 ns	1.01	1.075
1_cubic_oneshot/q10000	40172.7 ns	43528.5 ns	0.923	0.725
2_cubic_construct/g0100	1275.8 ns	1386.2 ns	0.92	0.917
2_cubic_construct/g1000	12716.0 ns	12722.9 ns	0.999	0.976
3_cubic_eval/q00001	20.9 ns	19.9 ns	1.05	1.005
3_cubic_eval/q00100	443.6 ns	442.0 ns	1.004	0.984
3_cubic_eval/q10000	39393.1 ns	42635.9 ns	0.924	0.905
4_linear_oneshot/q00001	24.6 ns	25.6 ns	0.961	0.967
4_linear_oneshot/q10000	18698.2 ns	18805.2 ns	0.994	0.991
5_linear_construct/g0100	34.8 ns	33.2 ns	1.048	0.821
5_linear_construct/g1000	241.0 ns	303.9 ns	0.793	0.849
6_linear_eval/q00001	10.1 ns	10.5 ns	0.962	0.99
6_linear_eval/q00100	196.6 ns	196.8 ns	0.999	0.986
6_linear_eval/q10000	18574.0 ns	18516.7 ns	1.003	0.991
7_cubic_range/scalar_query	8.3 ns	8.3 ns	1.0	0.981
7_cubic_vec/scalar_query	11.0 ns	11.3 ns	0.974	0.997
8_cubic_multi/construct_s001_q100	662.8 ns	647.8 ns	1.023	1.076
8_cubic_multi/construct_s010_q100	4453.2 ns	4460.8 ns	0.998	0.98
8_cubic_multi/construct_s100_q100	39645.7 ns	40208.4 ns	0.986	0.972
8_cubic_multi/eval_s001_q100	819.8 ns	827.4 ns	0.991	1.046
8_cubic_multi/eval_s010_q100	1837.3 ns	1813.8 ns	1.013	1.037
8_cubic_multi/eval_s010_q100_scalar_loop	2294.5 ns	2286.7 ns	1.003	0.983
8_cubic_multi/eval_s100_q100	11421.5 ns	11421.4 ns	1.0	0.996
8_cubic_multi/eval_s100_q100_scalar_loop	3363.4 ns	3352.3 ns	1.003	0.989
9_nd_oneshot/bicubic_2d	45646.0 ns	45192.7 ns	1.01	1.06
9_nd_oneshot/bilinear_2d	540.6 ns	539.8 ns	1.001	0.972
9_nd_oneshot/tricubic_3d	417419.3 ns	426524.4 ns	0.979	1.043
9_nd_oneshot/trilinear_3d	1031.9 ns	1040.0 ns	0.992	0.982

✅ 1 benchmark(s) were flagged but verified as CI noise after 10 re-run(s)

This comment was automatically generated by Benchmark workflow.

[codecov]

Codecov Report

❌ Patch coverage is 95.74468% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 96.56%. Comparing base (05c2b34) to head (4a3d7d3).

Files with missing lines	Patch %	Lines
ext/FastInterpolationsSymbolicsExt.jl	95.74%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #149      +/-   ##
==========================================
+ Coverage   96.40%   96.56%   +0.16%     
==========================================
  Files         148      148              
  Lines       12269    12272       +3     
==========================================
+ Hits        11828    11851      +23     
+ Misses        441      421      -20     

Files with missing lines	Coverage Δ
ext/FastInterpolationsSymbolicsExt.jl	95.83% <95.74%> (+44.72%)	⬆️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.