Cluster & Edit doesn't work anymore in OR 3.6.0

[RolfBly]

To Reproduce

Select any appropriate column, Edit cells, Cluster and edit; tick some boxes, select Merge and recluster.

Current Results

The first time, the same clusters re-appear, it hasn't processed any. A subsequent time, a message "You must check some Edit? checkboxes for your edits to be applied." appears, but some boxes are ticked.
The problem appears on a data set of just over 5000 rows, 6 columns.
On a small sample from my dataset, the problem is less severe. It processes 1 of 2 clusters selected.

Expected Behavior

In version 3.5.2 everything still works fine.

Screenshots

Versions

• Operating System: Windows 10 10.0.19044.1766
• Browser Version: Google Chrome 103.0.5060.134 (Official build) (64-bits)
• JRE or JDK Version: openjdk version "17.0.3" 2022-04-19
• OpenRefine: 3.6.0

[elebitzero]

I tried to reproduce this on Windows 10, Java17, OpenRefine 3.6 with a dataset with 88k rows and 40 columns, but it worked fine for me. I thought maybe it is a timing issue, so I used developer tools to introduce a 10 second lag time in network requests, but it still worked fine.

Do you see any errors in the JavaScript console, or on the command line where OpenRefine server is running? If you can share your dataset, we can try to reproduce the issue with the same data.

Operating System: Windows 10 Pro 10.0.19044.1826
Browser: Google Chrome Version 103.0.5060.134 (Official Build) (64-bit)
JRE or JDK Version: java 17.0.2 2022-01-18 LTS
OpenRefine: 3.6 downloaded 8/2/2022

[RolfBly]

Meanwhile,

• I've asked in the Google group how to share the data set privacy-proof - there's no formal way for it. I can zip it with password but how to get the password safely to you?
• I've tried the same dataset on a different machine with the same setup, problem appears there too
• I made a recording showing the problem
• I've asked my customer if they have any objection with me sharing the data set.

Just FYI, bear with me please.

Also, @wetneb, I noticed that too. Thanks for dotting the i.

[ostephens]

@RolfBly can you share the recording?
Also I notice that @elebitzero asked:

Do you see any errors in the JavaScript console, or on the command line where OpenRefine server is running?

Are you able to check either of these?

[RolfBly]

There are no errors in the console, it just stops as soon as the error message appears. The last line in the console is
14:58:11.688 [ compute-clusters_command] computed clusters [binning] in 196ms (29ms)
which is after a partially succesful merge & recluster.
Then you get a new list in the cluster & edit window.
Partially because a number of names that you did in the previous step, reappears, in other words it has not clustered them. So you re-select, hit Merge & re-cluster. At that point the error window appears. Nothing changes in the console. No lines added.

Unfortunately, my customer won't allow uploading the data or the screen recording here. If there's any other way of getting it to the right people, and to them only, please let me know.

That said, it's not that urgent. There is an easy workaround: just use version 3.5.2.

[ostephens]

@RolfBly thanks.

So far I've failed to recreate this locally - but that doesn't really mean anything beyond knowing that the problem doesn't occur in all circumstances.

I know this is asking more effort from you but just to try to see if we can recreate this on other datasets are you able to share step by step what you did (for example: is the Cluster option done from a facet or from the Edit Cells menu? Do you have any facets applied when you do the clustering? etc.) so we can just eliminate any variations in workflow from the issue? (I'm assuming that just writing down the steps is easier, but an alternative might be to share a screen recording with any confidential information blurred out - just trying to make sure I'm not missing any step when I try to recreate the issue locally)

[RolfBly]

Meanwhile, I've found an open source data set that I can hopefully reproduce the problem with. I'll have to do this in my spare time, so it may take a while.

There is one additional thing: my workspace is a Dropbox folder that gets auto-synced to the cloud, plus I've set the auto-save interval to 5 in openrefine.l4j.ini. This is because in the recent past, I've had a few important projects disappear or corrupted and I've never been able to find out why or how.

[ostephens]

What's the dataset @RolfBly - I can have a look as well

[RolfBly]

UFO_sightings.csv.zip
Here you go. The offending columns in my case has [name]|[iso-date] or just [name] where name to cluster and edit may be something like

Doe, john, simon
Doe, John Simon
DOE, John

It probably should work with location names too.

[ostephens]

Thanks @RolfBly Do you have a Text Facet displaying for the column (in the project Facet / Filter pane) while you are doing the clustering?

[RolfBly]

No.