[BUG]: missing tests in `ndarray/base/unary`

[kgryte]

We previously added tests to ndarray/base/nullary. This issue requests to extend those efforts to ndarray/base/unary.

Link: https://github.com/stdlib-js/stdlib/tree/develop/lib/node_modules/%40stdlib/ndarray/base/unary

Time estimate without AI: 6hrs
Time estimate with AI: 4-5hrs

[headlessNode]

@naterush I'm planning to work on this issue. Could you please assign this a label? Thanks!

[headlessNode]

PR: stdlib-js/stdlib#5821
Time till opening PR: 8hr 56mins
AI Model: Claude 3.7 Sonnet
Loom recording: Link

[headlessNode]

Notes:

• This issue was fairly simple in terms of the coding logic required. As mentioned in the issue, a similar implementation was available for reference.

• For this one, I experimented with multiple approaches:

  1. From test.0d.js to test.3d.js, I used Cursor Composer. It took more time than expected, likely because each test case had unique caveats. With more than 30 test cases, it was difficult to provide Composer with prompts that would generate everything correctly in one go. The time taken was due to two main factors: first, the time required for prompting and AI-generated code, and second, the time spent reviewing and making corrections to ensure everything was as expected.

  2. From test.4d.js to test.8d.js, I used the Cursor Tab. This approach was more efficient since I went through each test case individually from the start. When the Cursor Tab’s suggestion was appropriate, I used it; otherwise, I made manual adjustments.

  3. From test.9d.js to test.nd.js (the largest files), use of AI (both Cursor Composer and Tab) was minimal. Since the test cases were similar to the previous ones—requiring only a few tweaks to dimensions and additional cases—I copied the relevant cases from earlier files and modified them accordingly. This approach proved to be the most time efficient for this particular issue.

[headlessNode]

Post PR Review Notes:

• Additional 29 mins were spent on refactoring complex number cases.