Thanks, I'll take a look when I get a chance. May need to finish getting the test suite up-to-date so we can check if the changes are ok.

Can you give a brief example of the form of data that was slow? Just one or two properties is fine, not 10000+ :-) I've started setting up a benchmarking system and issues like this are useful target for auto-generating test data inputs.

I'd like to bump this as it fixes major issues we're facing internally, with slow framing on large JSON-LD files.

I will try to generate some artificial data (large amount of data, few properties).

Hi! Any plans to include this? We are also facing issues with framing large JSON-LD files.

@RinkeHoekstra I know it's been a while, but now that there are spec tests, this performance fix makes a Flattening and Transform test fail: https://github.com/digitalbazaar/pyld/actions/runs/21819124584/job/62947739897?pr=101

I have fixed the bugs in the perfomance version; the test-suite passes now. Not all performance gains have been kept, but there is still significant speedup.

@dvsrepo if after 5 years, you're still in the position of benchmarking this, please do :)

I ran a quick and dirty benchmark on the transformation tests:

Benchmark 1: master
  Time (mean ± σ):      2.226 s ±  0.435 s    [User: 0.631 s, System: 0.192 s]
  Range (min … max):    1.563 s …  2.790 s    10 runs
 
Benchmark 2: original-changes
  Time (mean ± σ):      1.905 s ±  0.310 s    [User: 0.548 s, System: 0.161 s]
  Range (min … max):    1.561 s …  2.437 s    10 runs
 
  Warning: Ignoring non-zero exit code.
 
Benchmark 3: current-changes
  Time (mean ± σ):      1.706 s ±  0.305 s    [User: 0.496 s, System: 0.146 s]
  Range (min … max):    1.557 s …  2.562 s    10 runs
 
  Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs.
 
Summary
  current-changes ran
    1.12 ± 0.27 times faster than original-changes
    1.30 ± 0.35 times faster than master