feat: Add unsafe array take and put operations

[mark-koch]

I'm reluctant to add this, but this should remove the need for people to use option-arrays.

Questions:

• Is put a good name?
• Should the names include an _unsafe suffix?

[codecov-commenter]

Codecov Report

:x: Patch coverage is 87.30159% with 8 lines in your changes missing coverage. Please review. :white_check_mark: Project coverage is 93.55%. Comparing base (495aba5) to head (1b3bc7b). :warning: Report is 3 commits behind head on main.

Files with missing lines	Patch %	Lines
guppylang/src/guppylang/std/array.py	61.90%	8 Missing :warning:

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1165      +/-   ##
==========================================
- Coverage   93.59%   93.55%   -0.04%     
==========================================
  Files         128      128              
  Lines       11451    11503      +52     
==========================================
+ Hits        10717    10762      +45     
- Misses        734      741       +7     

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

:rocket: New features to boost your workflow:

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

[ss2165]

This is waiting on an updated implementation using borrow array, which requires a "check" HUGR operation https://github.com/CQCL/hugr/issues/2569

[acl-cqc]

My main comment is that this makes the analysis for borrow-return "squashing" substantially harder (and/or less effective). See https://github.com/CQCL/tket2/pull/1159 . Doesn't mean we can't do this, but we may want to hold off.

UPDATE: I've reduced scope of that squashing meaning that it is now "safe" even given this PR, although https://github.com/CQCL/tket2/issues/1176 would still be useful (not essential).

[github-actions[bot]]

Bencher Report

Branch	feat/array-unsafe
Testbed	Linux

Click to view all benchmark results

Benchmark	Latency	Benchmark Result microseconds (µs) (Result Δ%)	Upper Boundary microseconds (µs) (Limit %)
tests/benchmarks/test_big_array.py::test_big_array_check	📈 view plot 🚷 view threshold	686,313.06 µs (-7.28%) Baseline: 740,166.66 µs	777,174.99 µs (88.31%)
tests/benchmarks/test_big_array.py::test_big_array_compile	📈 view plot 🚷 view threshold	1,636,192.38 µs (+3.93%) Baseline: 1,574,284.68 µs	1,652,998.92 µs (98.98%)
tests/benchmarks/test_big_array.py::test_big_array_executable	📈 view plot 🚷 view threshold	7,392,575.38 µs (+1.45%) Baseline: 7,287,270.12 µs	7,651,633.63 µs (96.61%)
tests/benchmarks/test_ctrl_flow.py::test_many_ctrl_flow_check	📈 view plot 🚷 view threshold	50,534.84 µs (-33.16%) Baseline: 75,604.05 µs	79,384.25 µs (63.66%)
tests/benchmarks/test_ctrl_flow.py::test_many_ctrl_flow_compile	📈 view plot 🚷 view threshold	90,098.21 µs (+1.61%) Baseline: 88,672.51 µs	93,106.13 µs (96.77%)
tests/benchmarks/test_ctrl_flow.py::test_many_ctrl_flow_executable	📈 view plot 🚷 view threshold	603,180.49 µs (+1.00%) Baseline: 597,223.46 µs	627,084.63 µs (96.19%)
tests/benchmarks/test_prelude.py::test_import_guppy	📈 view plot 🚷 view threshold	49.95 µs (-3.03%) Baseline: 51.51 µs	54.09 µs (92.36%)

🐰 View full continuous benchmarking report in Bencher

[github-actions[bot]]

Bencher Report

Branch	feat/array-unsafe
Testbed	Linux

Click to view all benchmark results

Benchmark	hugr_bytes	Benchmark Result bytes x 1e3 (Result Δ%)	Upper Boundary bytes x 1e3 (Limit %)	hugr_nodes	Benchmark Result nodes (Result Δ%)	Upper Boundary nodes (Limit %)
tests/benchmarks/test_big_array.py::test_big_array_compile	📈 view plot 🚷 view threshold	143.70 x 1e3 (0.00%) Baseline: 143.70 x 1e3	145.13 x 1e3 (99.01%)	📈 view plot 🚷 view threshold	6,590.00 (0.00%) Baseline: 6,590.00	6,655.90 (99.01%)
tests/benchmarks/test_ctrl_flow.py::test_many_ctrl_flow_compile	📈 view plot 🚷 view threshold	21.67 x 1e3 (0.00%) Baseline: 21.67 x 1e3	21.89 x 1e3 (99.01%)	📈 view plot 🚷 view threshold	606.00 (0.00%) Baseline: 606.00	612.06 (99.01%)

🐰 View full continuous benchmarking report in Bencher

[ss2165]

What should be the suffix for the panicking versions?

How about making put and take the panicking versions and make the fallible versions try_take and try_put ?

[mark-koch]

How about making put and take the panicking versions and make the fallible versions try_take and try_put ?

Actually, that sounds like a good option 👍