got 'finish' event before all pending jobs done

[jacobbubu]

When use it to implement a writeable stream, the 'finish' event fired before all of the pending jobs done.

[almost]

You're right that it should definitely wait until all transforms are complete before firing 'finish'. I'm trying to figure out a nice way of fixing this at the moment. If you had a test case showing the failure that'd be really helpful, even more so if it was in the form of a failing unit test :)

[serp-dev]

I also encountered such a problem. Here are small tests for dimontration:
serp-dev@f7b5248

[almost]

Thanks for this!

[YounGoat]

@jacobbubu Maybe it is triggered by duplicated invoking of callback?

[duanshiqiang]

This is a really annoying bug, it swallow us a whole afternoon to figure out why some chunk are missing. Hope it can be fixed asap..

[YounGoat]

How about trying this package? https://www.npmjs.com/package/oct <https://www.npmjs.com/package/oct> youngoat@163.com

…

On 22 Dec 2017, at 6:08 PM, Duan Shiqiang ***@***.***> wrote: This is a really annoying bug, it swallow us a whole afternoon to figure out why some chunk are missing. Hope it can be fixed asap.. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#7 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AH4XcbfnHQHK7BfIgSwycYHWwqq-dgVXks5tC3-ZgaJpZM4EE19p>.

[moander]

I tried this because I had the same issue using alternative package parallel-transform. Is it no real alternative to do get parallel streams?

mafintosh/parallel-transform#4

[Tapppi]

@almost This is due to using _flush to make sure all processing is done. That is called after the Writable side has already closed (and finish called). Node 8 added _final which is _flush but for the writing side, which allows to easily wait for all processing to complete without emitting finished.

An example fix is here mafintosh/parallel-transform#6, but that depends on readable-stream@^2.3.0 as that is the version that added Node 8 support. I think the only sensible way of surely fixing this here is to include readable-stream@^2.3.0 as a dependency to force through2 to also use a newer version (should work in any of the myriad dep tree cases?). Another possibility is to set both _final and _flush to fix it for anyone using non-ancient deps and fall back to the old behaviour otherwise.

Fixing this without _final involves monkey-patching end(), at least one dude used to do this before _final, but at least I don't feel like that 😂.