Summary:
Previously, ref/reflist columns were formatted entirely based on their visible column, since they received values from the visible or display columns rather than the actual row IDs. This creates `ReferenceFormatter` and `ReferenceListFormatter` which still delegate most of the formatting work to a visible column formatter but fix a few issues:
- ReferenceList columns now actually use the options (e.g. date format) of the visible column to format their elements. Previously they were formatted generically because the visible column formatter wasn't expecting a list.
- Invalid references aren't formatted with an `#Invalid Ref` prefix.
- When the ref column displays the Row ID, it doesn't have a visible or display column. Previously this led to the references being formatted as just numbers in most cases, with special code in the widget to display them like `Table1[2]`. Now they are consistently formatted in that style throughout.
Test Plan: Updated existing tests.
Reviewers: jarek
Reviewed By: jarek
Subscribers: dsagal
Differential Revision: https://phab.getgrist.com/D3212
Summary:
Formats lists as CSVs at the top level, so the list `["a", "b"]` gets formatted as `a,b`. Further nesting looks like JSON, with quotes around strings, which get doubled to escape them in the CSV. So the common case looks significantly nicer, but the rare case of nested arrays looks very weird and confusing.
There's also some smaller details about quotes and spaces to discuss if we're happy with the overall idea.
This is part of revamping type conversion and was discussed here: https://grist.quip.com/csqCAfx6KHt2#HOaADA7Q6NM
Test Plan: Updated several tests, need to confirm if we want this behaviour before continuing.
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D3208
Summary:
Reference/referencelist columns displaying date/datetime columns didn't show the formatting of that column, formatting them as ISO instead. One weird effect of this was that opening the editor suddenly changed the format because the editor formatted the dates correctly. You can see this in the checkin doc as an example.
This was discussed in https://grist.slack.com/archives/C0234CPPXPA/p1636482208111800. Here's the main point:
> both use the visible column formatter's formatAny. the editor uses the value from the visible column, which for a date column is a raw timestamp number. the cell display uses the value from the display column which is of type Any so the value is wrapped in a list starting with 'd'. the former gets formatted according to the formatting options, but the latter just gets formatted as ISO.
Probably a good solution to the broader problem is to ensure that the display column has the same type and widget options as the visible column. That seems potentially messy, so I did something easier: fix `DateFormatter` to accept encoded date/datetime objects. It still receives the correct widget options from the visible column as before but can handle the values from the display column. This might also have other uses in the future.
Test Plan:
- Fixed several tests which previously expected the buggy behaviour.
- Converted ValueFormatter.js tests to typescript and cleaned up the existing code slightly.
- Added tests for DateFormatter and DateTimeFormatter to the ValueFormatter test suite, which only tested numbers before.
Reviewers: dsagal
Reviewed By: dsagal
Subscribers: dsagal
Differential Revision: https://phab.getgrist.com/D3190
Summary:
Added ChoiceListParser capable of parsing JSON, CSVs, and other strings containing user-configured choices (e.g. separated by spaces)
I got a little carried away here. It works, and I can't think of any bugs, but it's complicated enough that there could be hidden edge cases or difficulties maintaining it in the future. The advantage of the current method is that it should work well for ambiguous or poorly formatted inputs, e.g. choices separated only by spaces or choices containing commas which are not escaped/quoted properly. The code can be vastly simplified if we don't try to support that and require that users paste proper JSON or CSVs.
Test Plan: Added a new file test/common/ChoiceListParser.ts with pure unit tests. Waiting for approval of the overall approach before adding to the nbrowser CopyPaste test.
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D3141
Summary:
Add ValueParser file, base class, and subclasses for column types. Only NumericParser is used for now.
Add valueParser field to ViewFieldRec.
Use valueParser when parsing pasted text data in Grid and Detail views.
Test Plan: Add test to nbrowser CopyPaste suite, copying into a numeric column with different currency and locale settings.
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D3082
Summary:
- Grist document has a associated "locale" setting that affects how currency is formatted.
- Currency selector for number format.
Test Plan: not done
Reviewers: dsagal
Reviewed By: dsagal
Subscribers: paulfitz
Differential Revision: https://phab.getgrist.com/D2977
Summary:
- Change RECORD's dates_as_str default to False.
- Reimplement objtype encode_object/decode_object with less machinery.
- Implement encoding of dicts (with string keys).
- Make lists and dicts encode values recursively.
- Implement encoding/decoding in the client
- Decode automatically in plugins' fetchSelectedTable/Record, with an option to skip.
Test Plan: Tested manually, not sure what tests may be affected yet.
Reviewers: paulfitz
Reviewed By: paulfitz
Differential Revision: https://phab.getgrist.com/D2593
Summary:
- Instead of sending an "UnmarshallableError" as an exception, introduce an
"Unmarshallable" type of value, represented as ['U', repr(value)]
- Unmarshallable values are rendered using a bluish text color, no longer a
pink background.
- Factor out ErrorDom to be simpler and cleaner.
- Add GristObjCode enum, and simplify related helpers.
- Use safe_repr() for when repr() itself fails
- Handle conversion errors using safe_repr() when str() fails
Test Plan: Added a test case based on a fixture covering a bunch of cases.
Reviewers: paulfitz
Reviewed By: paulfitz
Differential Revision: https://phab.getgrist.com/D2584
Summary: This moves enough server material into core to run a home server. The data engine is not yet incorporated (though in manual testing it works when ported).
Test Plan: existing tests pass
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D2552
