-
Notifications
You must be signed in to change notification settings - Fork 794
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
feat: Improve the syntax for conditions with multiple predicates #3427
Conversation
Returned type is only `core.SchemaBase` under one very specific case.
Appears to have been there since at least 5.0.0
Should cover all valid combinations, including that only one of `if_true`, `if_false` may be `str`.
Required for an annotation in later commit, but the import complicated the diff
- `expr.Expression` -> `_expr_core.Expression` fixes the type appearing as `Any` - `dict` -> `TypingDict[str, Any]` fixes the type appearing as `dict[Unknown, Unknown]` - Generally, the hierarchy of `Union`'s is allowing reuse and trailing docstrings, which I think improves readability. These will be used in `alt.condition` and related functions
Split out the `_PredicateType`/`_StatementType` parsing for reuse in `when-then-otherwise`.
Unsure if `expr` is still needed here as a side effect, but it is no longer needed internal to `api`. Purely a conservative step.
Allows the distinct steps to be reused
This operation was performed on every call, but was entirely static.
These need to be decoupled for the following cases: - Only `if_true` provided - multiple (`if_true`,`if_false`) - Ending of an `if_false` - Ending of an `if_true` Also adds a placeholder `TypeError`.
Resolves: > FAILED tests/utils/test_core.py::test_infer_encoding_types_with_condition - AssertionError: assert {'color': Col... value: 2})} == {'color': Col... value: 2})}
Will need updating with tests on `_Then`. Currently the wrapped `dict` should be a parseable input, but the rest of `altair` doesn't know what to do with `_Then` yet.
`list` fails on 3.8
…python>=3.12.0` Any objects using `__getattr__`, e.g. `Parameter` break on older versions [see PR](python/cpython#103034)
Within a recursive call, an `str` would always be checked for each case - despite the last always being false.
Somewhat duck-typing as `SchemaBase`.
Temporary solution while `when-then-otherwise` is not related to any other `altair` classes.
@mattijn All of #3427 (comment) is finished and ready for review. If you'd like to defer the docs part to another PR, I'd be happy with merging as-is. |
Thanks! Looks great! A few question for clarifications. I was confused about these two new tests: when_constraint = alt.when(Origin="Europe")
when_constraints = alt.when(
Name="Name_1",
Color="Green",
Age=25,
StartDate="2000-10-01"
)
expected_constraint = alt.datum.Origin == "Europe"
expected_constraints = (
(alt.datum.Name == "Name_1")
& (alt.datum.Color == "Green")
& (alt.datum.Age == 25)
& (alt.datum.StartDate == "2000-10-01")
) This is syntax I have not seen before. I noticed this is mentioned in the polars docs, that mentions this:
And
Do you know where this type of syntax comes from? Is it unique for polars or also used elsewhere? This only functions with a single |
@mattijn I'm confused also but for a different reason. altair/altair/vegalite/v5/api.py Line 1035 in 87cd9bd
It was also part of the original proposal #2759 (comment) I also listed it under the Stretch Tasks #3427 (comment) It utilises alt.when(**{'my column has spaces':4}) I would maybe note this in the User Guide as an option, but I'd probably lean on a different feature for those cases. Apologies, I'm doing this from my phone - but I'm pretty sure one of the later tests has this syntax of unpacking.
I'll update this tomorrow with a more complete response, but appreciate you checking this out so quickly |
Great, this docstring you refer to is very clear. Thanks for linking. It would be great if we can extend the documentation as well in a follow up PR on this new feature's scope to manage user expectations. Maybe without realizing, but there have been many new things in this PR! Thanks for all these efforts. I'm approving, but please comment when you think this PR can be merged! |
@mattijn thank you, I'm ready for merging Will absolutely be following up with another PR w/ more docs, especially examples on the methods (not just |
Thank you @dangotbanned! |
* fix: Use absolute imports explicitly Fixes the mypy issue referenced in `alt.__init__.py` * refactor: remove dynamic globals manipulation The results of `_populate_namespace` were only visible in a `IPython` environment * feat: Reimplement `expr` as a class w/ identicial semantics as the singleton - all `ConstExpression`s are now read-only properties stored in a metaclass - `expr` subclasses `ExprRef`, replacing the `__call__` w/ `__new__` - No instances of `expr` are ever created - All `FunctionExpression`s are now classmethods - Each `Expression` can now be referenced in autocomplete, without needing `IPython`/`Jupyter` - Docstrings have been reformatted to be compatible with `sphinx` - Broken links in docstrings fixed (e.g. for d3) - class-level doc for `expr`, with references & doctest * fix: use absolute imports in `test_expr` Previously relied on the dynamic globals * test(perf): rewrite `expr` tests to run in parallel Previously 2 tests, now 170+ * test: confirm `expr` constants stay constant * chore(typing): add ignores not flagged by mypy * docs: adds `alt.expr` to API Reference Uses the section heading proposed in #3427 (comment) Expands on #2904 * test: ensure `test_expr_consts_immutable` pattern works for all versions * docs: fix typos, regex artifacts, apply some `ruff` `D` rules * docs: add doc for `_ConstExpressionType` * test: make `expr` doctest testable I left this out initially, but by adding static `Parameter` names it is now possible * fix: re-run `generate-schema-wrapper` https://github.com/vega/altair/actions/runs/10005909777/job/27657526013?pr=3466 * refactor: Remove `expr` test dependency on constants Fixes #3466 (comment) * style: remove trailing commas in `@pytest.mark.parametrize`
Planned in vega#3427 (comment) Will allow for more reuse
* refactor: Rename and move `is_undefined` Planned in: https://github.com/vega/altair/pull/3480/files/419a4e944f231026322e2c4f137e7a3eb94735e8#r1679197993 - `api._is_undefined` -> `schemapi.is_undefined` - Added to `utils.__all__` * refactor: Rename and move `OneOrSeq` Planned in #3427 (comment) Will allow for more reuse
`field` proposed in vega#3239 (comment) `agg` was developed during vega#3427 (comment) as a solution to part of vega#3476
Remember discovering this during (#3427), but hadn't documented
Fixes #2759
Throughout this PR I frequently reference
polars.when
, and I want to make sure any reviewers have the origin handy.Tasks
Review
Investigate strange
SchemaValidationError
feat: Improve the syntax for conditions with multiple predicates #3427 (comment)_Conditions
->_Conditional
Investigate @mattijn proposed
str_as: Literal["value", "datum", "field", ...]
as an alternative tostr_as_lit
feat: Improve the syntax for conditions with multiple predicates #3427 (comment)str_as_lit=False
->str_as="shorthand"
Core
alt.condition
signature with overloadswhen-then-otherwise
syntaxotherwise
alt.condition
is usedthen
dfc14de
(#3427)when
function as entry-point_when
->when
38d425f
(#3427)when-then-otherwise
when
When.then
Then.when
Then.otherwise
ChainedWhen.then
when-then-otherwise
_Then
ending conditionsStretch
These would be nice to have and were demonstrated in #2759 (comment). Each idea comes from polars.when.
However, they need further discussion due to deviating from
alt.condition
's original behaviour.then
,otherwise
inalt.value
.38d425f
(#3427)when
asalt.datum.key.__eq__(value)
.38d425f
(#3427)alt.condition(**kwargs)
, for something other thanempty
:legend=None
.scale=alt.Scale(...)
.**kwargs
fromwhen
->then
,otherwise
.when
's predicate argument to be variadic, reducing all predicates as anAND
reduction.38d425f
(#3427)alt.condition
already accepts the result of this.when(predicate, **kwargs)
->when(*predicate, **kwargs)
doesn't change the meaning of arguments other thanpredicate
alt.condition