Handle mercator infinity projection at 90,-90 #1552
Conversation
|
Ref #1552 |
|
130 or so failures! Wow congrats @aashish24 😉 |
|
LGTM merging. 😉 |
|
Goes with CDAT/uvcdat-testdata#64 |
|
@doutriaux1 yes, we should make a rule to merge only those branches that can bring atleast 100+ test failures -:). Ready for review. One thing I noted (which I am not sure if caused by my branch), is that the text in the rendered plot overlaps slightly with the visualization. I am not sure if that is something related. |
| @@ -94,11 +94,27 @@ def putMaskOnVTKGrid(data, grid, actorColor=None, cellData=True, deep=True): | |||
| grid.GetPointData().AddArray(ghost) | |||
| return mapper | |||
|
|
|||
| def handleProjectionEdgeCases(projName, data): | |||
| # For mercator projection, latitude values of -90 or 90 | |||
| # transformation result in infinity values. We chose -85, 85 | |||
There was a problem hiding this comment.
Technically, 85.051129 is used because that makes the domain a square in projected coordinates.
https://en.wikipedia.org/wiki/Web_Mercator#Formulas
There was a problem hiding this comment.
@jbeezley sure, but matplotlib uses 85.0 and so as some other tools. For UV-CDAT since we scale the plot, which makes it a rectangle anyways but I am willing to change it to 85.051129 if that is preferred.
There was a problem hiding this comment.
No 85.0 is close enough for any practical purpose, I was just pointing out that it isn't a magic number chosen arbitrarily. It is derived for a reason.
There was a problem hiding this comment.
sure, yup, I knew -:) but you are right not everyone would know. On a related note, matplotlib has a threshold option as well that I would like to add at some point.
There was a problem hiding this comment.
On this, I would like to propose that VCS do not scale the viewport but that's something I need to bring up in some meeting. The idea is to fill the screen but to me anytime you use a projection, it really comes out distorted.
There was a problem hiding this comment.
👍 The aspect ratio of the projection should always be 1:1.
aashish24
force-pushed
the
fix_mercator
branch
from
c7961f7
to
a0166fe
Compare
|
Ref: #587 |
aashish24
force-pushed
the
fix_mercator
branch
from
a0166fe
to
8bd47f6
Compare
|
@doutriaux1 some of tests are failing because of the bad testdata branch. I repushed it. |
|
@aashish24 one test failed and no flake8. |
|
@aashish24 I'm fixing annie should be able to run the build soon after is merged #1551 |
|
@doutriaux1 all of the tests should pass now. |
|
@doutriaux1 ready to be merged.. all tests are passing. |
|
@doutriaux1 looks like RH6 is not pulling the right branch for test-data |
|
@aashish24 I wonder if ctest forces it back to master branch? I will take a look. The history show a successful switch to fix_mercator branch. |
|
@doutriaux1 I am going to merge this branch soon since all of the tests are passing now. |
|
Please keep it open I will merge it. I need to figure out why the bots don't check the correct branch |
|
@aashish24 seems like my bots are out of sync with the master branch, going to update them and retrigger the builds. Please hold on. |
|
@doutriaux1 ping!! |
|
let me try locallly |
|
@aashish24 I get a slight diff on my ubuntu: https://open.cdash.org/viewTest.php?onlyfailed&buildid=4017404 |
| x = vcs.init() | ||
| iso = x.createisofill() | ||
| iso.projection = "mercator" | ||
| x.plot(s(latitude=(-90, 90)), iso) |
There was a problem hiding this comment.
@aashish24 please use setbgdimension and plot in bg=1
There was a problem hiding this comment.
I suspect the slight diff on my ubunutu maybe due to default size being to small.
|
@doutriaux1 I have pushed changes as suggested. I may have to upload a new baseline as well. |
|
@aashish24 thanks |
|
@doutriaux1 ready for your review. |
|
I don't understand why the bots do not check the correct uvcdat-testdata, running the command locally works... |
Handle mercator infinity projection at 90,-90 ##bot##skip-commit
|
@doutriaux1 maybe get the |
Not ready for review yet.. just wanted to push for buildbot.