Search performance and search cancellation

[ertankucukoglu]

What problem does this solve?

• There are about 184000 files in database and a simple *.pas filter searches among them all. However, there is only 445 *.pas files in database.
• There is a risk of corrupting common temporary files when there are parallel searches.
• Cancelled search continues its own search operation and does not stop.
• Performance counters are not quite detailed.
• A lot of matching search like letter "a" is not stopping the PowerShell search at 500 files even if C part is only using the first 500 accepted result.

Proposed solution

I have below diff package against to the current master (hopefully it is still matching as master is quite frequently changing recently).
Total 5 files including new tests. Compiled and tested on my system.
cbm-main-60a2bbf.zip

Please feel free to use them as you like, or simply ignore if they break something else or does not feel right.

Thanks.

Alternatives considered

No response

Confirmations

• I searched existing issues and this is not a duplicate.

[DeusData]

@ertankucukoglu, thank you for profiling this and for preparing tests. We cannot review or apply source changes delivered only as an attached ZIP: it removes the normal GitHub diff, provenance, DCO, and per-commit review path, and the attachment is already drifting as main moves.

Please open standard pull requests against current main, one independently reviewable concern per PR. The natural slices are extension-filter narrowing, per-search temporary-file isolation, cancellation propagation, performance counters, and stopping the PowerShell producer at the accepted-result cap. Include the relevant regression test or measurement with each slice.

There is no need to maintain the ZIP for us. Focused GitHub PRs will let us review and credit the work normally. Thank you again for offering the improvements.