Avoid redundant deep cloning when unwinding.

[jarib]

AFAICT there's no reason to deep clone unless the unwind path has multiple levels. For large and deeply nested objects this should be a significant performance gain.

[coveralls]

Coverage increased (+0.05%) to 85.443% when pulling 8ba4101 on jarib:faster-unwind into 61d9808 on zemirco:master.

[coveralls]

Coverage decreased (-0.05%) to 85.35% when pulling b86dc24 on jarib:faster-unwind into 61d9808 on zemirco:master.

[juanjoDiaz]

This makes the code quite more unreadable and I'm not convinced about how much of a difference it makes from the performance point of view.

[jarib]

It makes a huge difference if you're working on large objects. It improves performance by 80% on the data I'm currently working on:

$ cat data/result.json | jq '.[0:5]' | time ~/src/json2csv/bin/json2csv.js --unwind 'hits' >| /dev/null
~/src/json2csv/bin/json2csv.js --unwind 'hits' >| /dev/null  5.03s user 0.11s system 89% cpu 5.743 total
$ cat data/result.json | jq '.[0:5]' | time ~/src/json2csv/bin/json2csv.js --unwind 'hits' >| /dev/null
~/src/json2csv/bin/json2csv.js --unwind 'hits' >| /dev/null  0.35s user 0.08s system 39% cpu 1.079 total

[jarib]

I agree the readability isn't great though. I can try to clean it up a bit.

[jarib]

I think this is the best I can do right now :)

[knownasilya]

Thanks for submitting and working with us on getting it mergeable 👍

[knownasilya]

Published in v4.1.0