Closed doredesign closed 5 years ago
Hello @denisoster and thank you for the contribution 🛠
I've checked this one. The described problem really exists in test scenario.
But I want to check this in real application, next weak. And consult with @Fivell
@workgena Sounds good. 👍
I've checked the solution manually. The problem and solution confirmed.
@Fivell what do you think about merging this PR, as for me - it is ready.
@workgena, looks reasonable for me.
Version 4.1.1 has been published, it includes this fix. Gratitude to @doredesign 🎉
Currently the
use_indexes
are set correctly only for the first batch.You may be wondering why I re-ordered the columns in authors.csv. I did this because it was the best way I could find to get a good failing test for my use case. I could not test removing the
name
because that column is required. I couldn't test removing thelast_name
column because that column needs to be unique and validation would fail for the second batch because the first batch (and record) would have aNULL
value forlast_name
. I also could not use thedate
column in its former order because the bug only manifests when columns other than the last one(s) are sliced.