Should import CSV continue on index error?

[RafaPolit]

What should be the expected behavior if you import a CSV that would produce errors (in this particular case, indexation errors)?

Scenario A

• It stops as soon as it finds an error and reports the error for that particular field

Advantages:

• If it is a long import, this could prevent a long wait if this is a root error (ie. one property is wrongly named). You wouldn't have to wait for potentially hours waiting to just know that it didn't work

Scenario B

• It goes through the entire process and reports at the end which lines failed

Advantages:

• If there were only a few errors and apart from each other, you just need to fix the few errors instead of repeating the process.

I'm not sure. Perhaps an option on import? "Continue on error", leave the end user the decision?

At any rate, after #2931 and even after #2993 the process is stopping on first error. We should decide if this is what we want.

@txau @simonfossom

[txau]

@RafaPolit I guess it depends on the development effort. If we have to pick a good default, I'd say halt on errors has more advantages than the other one. Of course it is situational and there are different use cases. If implementing the option for users to choose isnot that much extra development, I'd go for that one with still "halt on errors" as the default.

[simonfossom]

Hmm... I would say that this is one of those We don't know situations where we should ask if it doesn't make a difference for us.

If we halt on error user must redo import from that line further. But there might be more errors in the import. So we make him redo many times over.

If we give a log of lines that failed and why, he can fix all errors at once and be done.

If we encourage user to do bulk actions we shouldn't waste his time. I would go with scenario B

[txau]

@simonfossom the decision was made considering development effort over UX, still the discussion is valid. I think the process could be more fine grained with some options:

• Prevalidate metadata without importing. This will already catch many potential issues without importing anything. This is crucial since "not halting on errors" forces the users to delete what has been imported.
• Be able to cancel the import process from the UI: #3423
• User can choose on whether cancel on error or continue on error.