allow unsetting options in PlotDataItem and PlotCurveItem by NilsNemitz · Pull Request #2041 · pyqtgraph/pyqtgraph

NilsNemitz · 2021-10-24T16:40:13Z

This is a quick attempt to address #2039: Various plotting parameters can be set, but cannot be unset, especially when going through the PlotDataItem interfaces.

The origin of the problem is that much of the code implicitly assumes that a new element is initialized from zero, and there are explicit checks that cancel out of the update early when parameters are None.

This problem is present in both PlotDataItem and PlotCurveItem, but the faulty shortcut checks are particularly used in the setData() methods. I think the problem is worse when going through a PlotDataItem, because this tends to call PlotCurveItem.setData() with any changes,

I made a quick pass over the code and removed problematic shortcut checks for None, as well as unconditional mkBrush / mkPen calls that prevented setting these to explicit None values.

The tests seem to pass, and this seems to address the test cases described in #2039.
But I have probably missed a few things, so some additional testing and feedback would be appreciated!

@pijyoi , do you immediately see anything that this misses?

Closes #2039 ... once it works.

pijyoi · 2021-10-24T19:05:43Z

(deleted, discussion in #2039)

pijyoi · 2021-10-29T00:32:51Z

So almost all the options are "sticky" options. If you set an option in one call of setData, it will persist.
This supports usage such as:

item = pg.PlotCurveItem(pen='w', brush='b', connect='finite')
item.setData(x, y)

The exception is skipFiniteCheck which gets disabled if you don't specify it in setData. I don't know if it should be made consistent with the rest.

j9ac9k · 2021-10-29T01:49:48Z

I can be talked into skipFiniteCheck going either way. The first use case was intended to implement the "safest" way, to minimize the likelihood of that check accidentally being skipped,

I'm still partial to that mechanism; but I'm not sure we should have one option behave differently from the others.

j9ac9k · 2021-10-30T05:23:51Z

@NilsNemitz do you think skipFiniteCheck should be a sticky option along w/ the others?

NilsNemitz · 2021-10-30T06:38:00Z

Hmm. So far, I always read "skipFiniteCheck" as a hint for just the current dataset.

But there is an argument to be made for turning it sticky:
In many cases, your data source will never produce non-finite values. Being forced to set the hint for each setData() call is needless boilerplate then.
Doing that is only useful if

non-finite values occur only occasionally
the user code already knows when they are present.

And where that rare case applies, a sticky flag can still be updated with every call. So I think a sticky flag might be better.

NilsNemitz · 2021-10-31T08:04:53Z

pyqtgraph/graphicsItems/PlotCurveItem.py

-
-        self.opts['skipFiniteCheck'] = kargs.get('skipFiniteCheck', False)
+        if 'skipFiniteCheck' in kargs:
+            self.opts['skipFiniteCheck'] = kargs['skipFiniteCheck']


This updates skipFiniteCheck only on request. A default False value will be used if it is never explicitly set.

NilsNemitz · 2021-10-31T08:10:06Z

The latest commit should make skipFiniteCheck a sticky flag that can be set once, and will still be effective during later setData() calls. It also adds a setSkipFiniteTest() method (for both PlotDataItem and PlotCurveItem) to allow setting/unsetting this outside of setData().

I have updated the documentation for the logic and effect, and also added some text that explains the recently added handling of wide lines.

I would also suggest adding a final one-year expiration period for the long-deprecated options stepMode=True, decimate and identical. I have added that for now. The main motivation would be to keep the code simple. If there is a reason to keep these mostly non-functional parameters, please let me know!

pijyoi · 2021-10-31T08:25:45Z

pyqtgraph/graphicsItems/PlotDataItem.py

+        require merging the entire plot into a single entity before the alpha value can be applied. For plots with more
+        than a few hundred points, this can result in excessive slowdown.
+
+        Since version 0.12.3, this slowdown is automatically avoided by an algorithm that draws line segments


Since 0.12.4 (or whichever version comes after 0.12.3)

j9ac9k · 2021-11-06T06:01:16Z

pyqtgraph/graphicsItems/PlotCurveItem.py

@@ -496,7 +515,10 @@ def updateData(self, *args, **kargs):
        if self.opts['stepMode'] in ("center", True):  ## check against True for backwards compatibility
            if self.opts['stepMode'] is True:
                import warnings


I know you didn't just add this, but would you mind removing this import here given the static code checker is complaining about it?

pijyoi · 2021-11-07T09:12:13Z

pyqtgraph/graphicsItems/PlotCurveItem.py

+                        by line segments. 
+
+                            | 'all' (default) indicates full connection. 
+                            | 'pairs' omits even-numbered segments. 


The previous docs say that even-numbered segments are drawn, presumably counting from zero.

Hmm. If you naturally start counting from zero, then neither of these descriptions seems particularly helpful.
I would start counting segments from one, and found the previous text confusing.

How about:
" 'pairs' draws one separate line segment for each two points given, starting with the first point. " ?

I suppose 'pairs' draws one separate line segment for each two points given. would be good enough. Don't think there is any doubt about it starting from the first point.

Agreed, and changed.

j9ac9k · 2021-11-18T04:41:10Z

Thanks @NilsNemitz you clearly have been busy ...love the documentation updates too 👍🏻 merging.

…2041) * allow unsetting parameters in PlotDataItem and PlotCurveItem * sticky skipFiniteCheck and documentation updates * can't count * removed duplicate import of wanrings * minor documentation clean-up for PlotCurveItem * clarify pairs parameters

allow unsetting parameters in PlotDataItem and PlotCurveItem

2787c45

NilsNemitz mentioned this pull request Oct 24, 2021

some PlotCurveItem options not resettable by PlotDataItem #2039

Closed

NilsNemitz added 2 commits October 31, 2021 16:00

Merge remote-tracking branch 'upstream/master' into PDI_parameter_pass

c611ba4

sticky skipFiniteCheck and documentation updates

7421c8e

NilsNemitz commented Oct 31, 2021

View reviewed changes

pijyoi reviewed Oct 31, 2021

View reviewed changes

can't count

9fca6ef

j9ac9k reviewed Nov 6, 2021

View reviewed changes

NilsNemitz added 2 commits November 7, 2021 14:07

removed duplicate import of wanrings

9a05738

minor documentation clean-up for PlotCurveItem

1b6d2b2

pijyoi reviewed Nov 7, 2021

View reviewed changes

clarify pairs parameters

44aa08e

j9ac9k merged commit 21d77db into pyqtgraph:master Nov 18, 2021

Uh oh!

Conversation

NilsNemitz commented Oct 24, 2021

Uh oh!

pijyoi commented Oct 24, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pijyoi commented Oct 29, 2021

Uh oh!

j9ac9k commented Oct 29, 2021

Uh oh!

j9ac9k commented Oct 30, 2021

Uh oh!

NilsNemitz commented Oct 30, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

NilsNemitz Oct 31, 2021

Choose a reason for hiding this comment

Uh oh!

NilsNemitz commented Oct 31, 2021

Uh oh!

pijyoi Oct 31, 2021

Choose a reason for hiding this comment

Uh oh!

j9ac9k Nov 6, 2021

Choose a reason for hiding this comment

Uh oh!

pijyoi Nov 7, 2021

Choose a reason for hiding this comment

Uh oh!

NilsNemitz Nov 7, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

pijyoi Nov 7, 2021

Choose a reason for hiding this comment

Uh oh!

NilsNemitz Nov 7, 2021

Choose a reason for hiding this comment

Uh oh!

j9ac9k commented Nov 18, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

pijyoi commented Oct 24, 2021 •

edited

Loading

NilsNemitz commented Oct 30, 2021 •

edited

Loading

NilsNemitz Nov 7, 2021 •

edited

Loading