Issues spotted in 2.6RC1

Joe Wicentowski

unread,

Sep 4, 2015, 2:05:56 PM9/4/15

to openr...@googlegroups.com

Hi all,

I've been really enjoying OpenRefine - thanks to the developers for all of their hard work!

Being new to the project, I'm not quite sure what the preferred way is to report issues to the developers - here vs. the developer list vs. directly on the GitHub issue tracker. Some issues I'd like to report include:

- In the "Cluster & Edit" window, the "Browse this Cluster" link opens a window with zero rows if any of the candidates the cluster contain characters like curly/typographic quotation marks, e.g., “. (I'm guessing this has to do with character escaping.)

- Inconsistent selection behavior with "Cluster & Edit": In the "Cluster & Edit" window, if I select a checkbox (to indicate that I want to merge entries) and select merge, the merge works as expected. Or if instead I click on one of the entries (to change the value to be used for merging) and then select merge, the merge works as expected. But if I select a checkbox AND click on one of the entries, I get an error dialog, stating: "The page at localhost:3333 says: You must check some Edit? checkboxes for your edits to be applied." Only by carefully *un*checking the checkbox will the merge proceed as intended. Alternatively, re-clicking on one of the entries will allow the merge to proceed. Clicking on an entry should only re-affirm the checkbox's checked status; it should not un-select the checkbox or cause a hidden selection toggle.

- Unexpected behavior when using "Apply to all identical cells" on blank cells: When I edit a blank/null cell via "Edit this cell" and select "Apply to all identical cells", neither the original cell nor the other blank/null cells are updated. The yellow-highlighted status message at the top-center of the window says merely, "Mass edit 0 cells in column title". At first I didn't notice that data wasn't updating, and as a result I lost work. I can see now that "Apply to all identical cells" only seems to work on cells that are not blank/null.

- Edits made via "Edit this cell" occasionally fail to apply: In some circumstances that I'm not able to consistently reproduce, changes made a cell via "Edit this cell" and "Apply" are not applied.

- Results returned via reconciliation service are not sorted by score: In building my own reconciliation service, I noticed that OR displays the results in the order returned by the remote API, rather than sorting them by the score object. For example, a query for "McCarthy, Joseph M." returns 3 results, which are displayed in the following order, even though the 3rd match has the highest score:

{"q0": {"result": [

{

"name": "McCarthy, Joseph R.",

"score": 6.4928436,

"type": ["person"],

"id": "6603"

},

{

"name": "McCarthy, Senator Joseph R.",

"score": 6.4928436,

"type": ["person"],

"id": "3558"

},

{

"name": "McCarthy, Joseph M.",

"score": 11.757351,

"type": ["person"],

"id": "6536"

}

]}}

OR should use the score value rather than the order of objects sent by the API (in particular since JSON lacks a notion of document ordering that either server or client can depend on).

I'm not sure if any of these are known issues or if they've been addressed in the develop branch since the release of RC1, but if not, I'd be happy to register them either on the developer list or as an issue on GitHub. Could someone please guide me as to the best route for reporting these issues?

Thanks very much for any suggestions you might have,

Joe

Martin Magdinier

unread,

Sep 4, 2015, 4:37:50 PM9/4/15

to openr...@googlegroups.com

Joe,

Thank you for this detailed report. Much appreciated to make OpenRefine stronger.
Best way is to open new issue in Github so the dev team can answer to each one of them.

Regards
Martin

--
You received this message because you are subscribed to the Google Groups "OpenRefine" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openrefine+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.