DM-48256: Fix RunTimeError 'Record data is not contiguous in memory.' when running subtraction #369
Conversation
parejkoj
left a comment
parejkoj
left a comment
There was a problem hiding this comment.
Could we get a unittest that confirms the bug is fixed? Because as coded, this wouldn't fix it (and I'm now not sure what the underlying bug truly is anyway!).
| xv = (np.rint(sources.getX() - mask.getX0())).astype(int) | ||
| yv = (np.rint(sources.getY() - mask.getY0())).astype(int) | ||
|
|
||
| # The source catalog may not be contiguous. |
There was a problem hiding this comment.
I don't think this comment is necessary: the lines below say the same thing.
| yv = (np.rint(sources.getY() - mask.getY0())).astype(int) | ||
|
|
||
| # The source catalog may not be contiguous. | ||
| if not sources.isContiguous(): |
There was a problem hiding this comment.
I'm pretty sure this is too late to fix the problem, since the error report has it coming from the xv=... line above. When I said line 854, it must have been on a different version of the file. What I had meant was just after the call to self.sourceSelector.selectSources in _sourceSelector(). Except _checkMask is called on sources, not selected, so I don't see how it could fail in this way, unless the input catalog is broken somehow?
There was a problem hiding this comment.
I put the putative fix in _checkMask because it operates on the source table in question, but you're probably right that this would best be handled earlier.
mrawls
force-pushed
the
tickets/DM-48256
branch
from
209be87 to
4dbfa54
Compare
|
I'd like to add a unit test but I'm not sure how to make a non-contiguous source catalog on purpose. I think the test would best go here; I can work on this tomorrow. |
|
I think we need to find out how this was triggered in this first place. I posted a followup question on the ticket, because it doesn't seem possible to have failed with a non-contiguous, unmodified catalog. |
No description provided.