refine auto bar mark determination #1801
Conversation
|
Ah, foo. Looks like I’ve broken the autoRectColorReducer test… I’ll need to investigate. |
mbostock
force-pushed
the
mbostock/auto-bar
branch
from
2dcc179 to
8f8140e
Compare
|
Fix was easy! I just needed to make colorReduce take priority over xReduce and yReduce. |
There was a problem hiding this comment.
Nice!! So I guess isOrdinalReduce can be effectively inlined because you made the reducer the top-level switch?
Nice that it no longer depends on xZero / yZero, given the intuition that "bar" implies zero-ness. (Maybe explicit zero-ness let us break ties between barX and barY? But I can't think of a case.)
Played around and didn‘t see anything unexpected.
Test seems good. None of our sample data seems to naturally hit this case. I feel like it’s more about small-N stuff. Like one revenue number for each of three business segments for each of the last five years.
* refine auto bar mark determination * add another test
* refine auto bar mark determination * add another test
Fixes #1800. The logic is quite complicated — but I think it helped to start by looking at the reducers, since that determines whether there is a bin or group transform.