-
-
Notifications
You must be signed in to change notification settings - Fork 7.9k
[Bug]: LogFormatter minor ticks with minor_thresholds
of (0,0) does not behave as documented
#25896
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
I think the fix here is diff --git i/lib/matplotlib/ticker.py w/lib/matplotlib/ticker.py
index 4c33e8c9af..5cb1ca6746 100644
--- i/lib/matplotlib/ticker.py
+++ w/lib/matplotlib/ticker.py
@@ -964,7 +964,7 @@ class LogFormatter(Formatter):
# It's probably a colorbar with
# a format kwarg setting a LogFormatter in the manner
# that worked with 1.5.x, but that doesn't work now.
- self._sublabels = {1} # label powers of base
+ self._sublabels = np.array([1]) # label powers of base
return
b = self._base
@@ -995,17 +995,16 @@ class LogFormatter(Formatter):
if numticks > self.minor_thresholds[0]:
# Label only bases
- self._sublabels = {1}
+ self._sublabels = np.array([1])
elif numdec > self.minor_thresholds[1]:
# Add labels between bases at log-spaced coefficients;
# include base powers in case the locations include
# "major" and "minor" points, as in colorbar.
- c = np.geomspace(1, b, int(b)//2 + 1)
- self._sublabels = set(np.round(c))
+ self._sublabels = np.geomspace(1, b, int(b)//2 + 1).round()
# For base 10, this yields (1, 2, 3, 4, 6, 10).
else:
# Label all integer multiples of base**n.
- self._sublabels = set(np.arange(1, b + 1))
+ self._sublabels = np.arange(1, b + 1)
def _num_to_string(self, x, vmin, vmax):
return self._pprint_val(x, vmax - vmin) if 1 <= x <= 10000 else f"{x:1.0e}"
@@ -1021,11 +1020,10 @@ class LogFormatter(Formatter):
fx = math.log(x) / math.log(b)
is_x_decade = _is_close_to_int(fx)
exponent = round(fx) if is_x_decade else np.floor(fx)
- coeff = round(b ** (fx - exponent))
if self.labelOnlyBase and not is_x_decade:
return ''
- if self._sublabels is not None and coeff not in self._sublabels:
+ if self._sublabels is not None and not np.isin(x, self._sublabels * b**exponent):
return ''
vmin, vmax = self.axis.get_view_interval() i.e. to detect whether if a minor tick is at a (sublabel * b**expoenent value) (e.g. 2*10^1 = 20), just compute all these values, rather than trying to round() the mantissa (this is wrong if we have subticks coming from an AutoLocator in the case of #12865. Tests & PR left as an exercise. |
Actually the fix above is likely wrong because it would break #12865 (we actually do want these intermediate ticks to be labeled). I'm not entirely sure what the clean fix would be, but the patch at #29698 (comment) still seems reasonable to me (even though it seems to be mostly working around the underlying issue). |
Bug summary
In investigating #25894 I discovered that despite being documented as:
It actually only removes some minor tick labels (e.g. with base of 10: 11-14 will get labels, but 16-19 (15 is edge case, dependent on floating point rounding, 1.4, or 140 also similar))
Code for reproduction
Actual outcome
Expected outcome
Additional information
Only really affects plots that are zoomed in to much less than one decade, which arguably doesn't make sense to use a log scale, but still seems incongruous with the documented "this turns off minor ticks".
Has to do with when
coeff
rounds down to1
in:matplotlib/lib/matplotlib/ticker.py
Lines 980 to 988 in f051d94
I'm not quite sure what the proper solution is, perhaps making
self._sublabels=set()
instead?Operating system
Linux
Matplotlib Version
3.8 (main)
Matplotlib Backend
No response
Python version
3.11
Jupyter version
No response
Installation
git checkout
The text was updated successfully, but these errors were encountered: