Decrease filter step size for small domains #958
Conversation
Otherwise the detected filter step size may be larger than the domain, which results in not being able to use the filter. This is applicable when working with small percentages (<0.1%). While changing this I also noticed this will treat negative numbers as very small; I changed `getNumericStepSize` to compute the step size based on absolute value. Signed-off-by: Isaac Brodsky <isaac@isaacbrodsky.com>
isaacbrodsky
force-pushed
the
filter-step-size
branch
from
06b0d48 to
a1c6d07
Compare
src/utils/filter-utils.js
Outdated
| } else if (diff <= 1) { | ||
| // Try to get at least 100 steps - and keep the step size below that of | ||
| // the (diff > 1) case. | ||
| const x = diff / 100; |
There was a problem hiding this comment.
can you make this min step to be 1000? currently when 1 < diff < 3 step = 0.001, and when diff = 1, step = 0.01
There was a problem hiding this comment.
Sure, updated. I also updated tests based on the values I saw
edit: I still have an issue in CI (https://travis-ci.com/keplergl/kepler.gl/jobs/290743951) but I don't get that error locally - any ideas?
Signed-off-by: Isaac Brodsky <isaac@isaacbrodsky.com>
isaacbrodsky
force-pushed
the
filter-step-size
branch
from
0736886 to
721163e
Compare
|
From the Travis CI runs, it looks like the results are not consistent between Node 8 and Node 10. Is there an assertion utility for asserting "floating point equal within some epsilon"? |
|
you can try use |
Signed-off-by: Isaac Brodsky <isaac@isaacbrodsky.com>
isaacbrodsky
force-pushed
the
filter-step-size
branch
from
8c4f22f to
a1854f6
Compare
|
Seems the tests are passing now although I didn't update them to use |
This reverts commit f35c403.
Otherwise the detected filter step size may be larger than the domain,
which results in not being able to use the filter. This is applicable
when working with small percentages (<0.1%).
While changing this I also noticed this will treat negative numbers as
very small; I changed
getNumericStepSizeto compute the step sizebased on absolute value. (Not sure the UI fully supports selecting negative
filter ranges, seemed like it wouldn't let me select them via mouse.)