Minimizing this would be helpful because the following works fine:

import matplotlib.pyplot as plt
import numpy as np
f, ax = plt.subplots()
times = np.arange('2005-02', '2005-03', dtype='datetime64[D]')
y = np.linspace(0, 10, len(times))
y[3] = 1
y[10] = 1
y[21] = 1
ax.scatter(times, [x if x != 1 else np.nan for x in y], color='b', marker='o')
plt.show()

I managed to find a minimal example:

import numpy as np
import pylab as plt

times = np.arange('2005-02', '2005-03', dtype='datetime64[D]')

y = np.random.random(size=len(times))
y[0] = np.nan
plt.scatter(times, y)

plt.show()

It fails when the first number on the y values in the scatter is a nan, and the x values are dates.

Looks like

diff --git i/lib/matplotlib/units.py w/lib/matplotlib/units.py
index b4677bdd3..d77ed8c5b 100644
--- i/lib/matplotlib/units.py
+++ w/lib/matplotlib/units.py
@@ -123,34 +123,38 @@ class ConversionInterface:
     @staticmethod
     def is_numlike(x):
         """
         The Matplotlib datalim, autoscaling, locators etc work with scalars
         which are the units converted to floats given the current unit.  The
         converter may be passed these floats, or arrays of them, even when
         units are set.
         """
         if np.iterable(x):
             for thisx in x:
+                if thisx is ma.masked:
+                    continue
                 return isinstance(thisx, Number)
         else:
             return isinstance(x, Number)
 
     @staticmethod
     def is_natively_supported(x):
         """
         Return whether *x* is of a type that Matplotlib natively supports or
         *x* is array of objects of such types.
         """
         # Matplotlib natively supports all number types except Decimal
         if np.iterable(x):
             # Assume lists are homogeneous as other functions in unit system
             for thisx in x:
+                if thisx is ma.masked:
+                    continue
                 return (isinstance(thisx, Number) and
                         not isinstance(thisx, Decimal))
         else:
             return isinstance(x, Number) and not isinstance(x, Decimal)
 
 
 class DecimalConverter(ConversionInterface):
     """
     Converter for decimal.Decimal data to float.
     """

fixes the issue as of master, which makes sense. (No tests provided :p)

There are some more questions, e.g. if thisx is ma.masked, do we continue iterating (as this patch does), or do we just unmask the array to start with and expect the unmasked values to have correct type?

Also only the change to is_natively_supported is necessary here, but I think the change to is_numlike is likely necessary in other cases too.

This works fine for plot, so yet again its a case of scatter being special. Why is the y data being masked by scatter in the first place?

This is part of the work on scatter to make it handle updates to datasets that initially had some missing data more gracefully. With plot we always have exactly 2 vectors to work with, with scatter we have up to 4 (x, y, size, color).

The bug is still there when everything is NaN:

import numpy as np
import pylab as plt

times = np.arange('2005-02', '2005-03', dtype='datetime64[D]')

y = np.random.random(size=len(times))
y[:] = np.nan
plt.scatter(times, y)

plt.show()

What exactly would we expect to happen if all data points are NaN?

Similar open issues for all-NaN: #14439, #14124

I would expect it to behave as it does for individual nans: not plot anything.

Background of my motivation: I am using NaNs to mask away data points that I want to have in different alphas. Usually, every plot contains both, so it works, but today I found one where all of them were of the same kind, and the other was fully masked. I have added a check to skip it alltogether, but it used to work on earlier versions of Matplotlib, so it should still work.

fig,ax= plt.subplots(figsize=(15,5))

ax.plot(consum_test.index,consum_test["PJME_MW"],label="Actual")
ax.plot(consum_test.index,consum_test["Prediction"],alpha=.5,zorder=10,label="Predicted")

consum = consum_test["PJME_MW"]
pred = consum_test["Prediction"]

plt.fill_between(consum_test.index, consum,pred, facecolor="green", alpha=.2,label="Difference")
from datetime import datetime

ax.set_ylim(25000, 45000)
ax.set_xbound(lower="2017-03-12 00:00:00", upper="2017-03-12 23:30:00")

plt.xlabel("Date", alpha=0.75, weight="bold")
plt.ylabel("Consumption", alpha=0.75, weight="bold")

plt.xticks(alpha=0.75,weight="bold", fontsize=11)
plt.yticks(alpha=0.75,weight="bold", fontsize=11)

plt.title("Period with the worst hourly prediction", alpha=0.75, weight="bold", fontsize=15, pad=10, loc="left")
plt.legend()

=========================================================================

this is my code and giving error

IndexError Traceback (most recent call last)
~\Anaconda3\lib\site-packages\matplotlib\axis.py in convert_units(self, x)
1522 try:
-> 1523 ret = self.converter.convert(x, self.units, self)
1524 except Exception as e:

~\Anaconda3\lib\site-packages\matplotlib\dates.py in convert(value, unit, axis)
1895 """
-> 1896 return date2num(value)
1897

~\Anaconda3\lib\site-packages\matplotlib\dates.py in date2num(d)
424 return d
--> 425 tzi = getattr(d[0], 'tzinfo', None)
426 if tzi is not None:

IndexError: too many indices for array: array is 0-dimensional, but 1 were indexed

The above exception was the direct cause of the following exception:

ConversionError Traceback (most recent call last)
in
12
13 ax.set_ylim(25000, 45000)
---> 14 ax.set_xbound(lower="2017-03-12 00:00:00", upper="2017-03-12 23:30:00")
15
16 plt.xlabel("Date", alpha=0.75, weight="bold")

~\Anaconda3\lib\site-packages\matplotlib\axes_base.py in set_xbound(self, lower, upper)
3168 upper = old_upper
3169
-> 3170 self.set_xlim(sorted((lower, upper),
3171 reverse=bool(self.xaxis_inverted())),
3172 auto=None)

~\Anaconda3\lib\site-packages\matplotlib\axes_base.py in set_xlim(self, left, right, emit, auto, xmin, xmax)
3292
3293 self._process_unit_info(xdata=(left, right))
-> 3294 left = self._validate_converted_limits(left, self.convert_xunits)
3295 right = self._validate_converted_limits(right, self.convert_xunits)
3296

~\Anaconda3\lib\site-packages\matplotlib\axes_base.py in _validate_converted_limits(self, limit, convert)
3206 """
3207 if limit is not None:
-> 3208 converted_limit = convert(limit)
3209 if (isinstance(converted_limit, Real)
3210 and not np.isfinite(converted_limit)):

~\Anaconda3\lib\site-packages\matplotlib\artist.py in convert_xunits(self, x)
173 if ax is None or ax.xaxis is None:
174 return x
--> 175 return ax.xaxis.convert_units(x)
176
177 def convert_yunits(self, y):

~\Anaconda3\lib\site-packages\matplotlib\axis.py in convert_units(self, x)
1523 ret = self.converter.convert(x, self.units, self)
1524 except Exception as e:
-> 1525 raise munits.ConversionError('Failed to convert value(s) to axis '
1526 f'units: {x!r}') from e
1527 return ret

ConversionError: Failed to convert value(s) to axis units: '2017-03-12 00:00:00'

how i can solve plz help

@wardafiaz please take user help to https://discourse.matplotlib.org! (but "2017-03-12 00:00:00" is a string. Convert to a date to use it as an limit).

Could anyone provide solution to this problem? thanks in advance

This issue is closed. Can you post a new issue with a self-contained example? Thanks...

Hey @Adi07-Nerd, where did you found a solution to your problem. I have expended a good couple of hours already trying to fix the issue below:

for dataframe in to_merge:
    
    names = dataframe.columns
    name = names[2]
    x = dataframe['Timestamp']
    y = dataframe['Manual_Count'].fillna(0)
    y[0] = np.nan
    z = dataframe.iloc[:,2].fillna(0)
    z[0] = np.nan
    plt.bar(x, [x if x != 1 else np.nan for x in y], label='Manual Count', width=0.2, color='cyan')
    plt.bar(x, [x if x != 1 else np.nan for x in z], label= 'Sensor data', width=0.2, color='indianred')
    plt.legend()
    ax = plt.gca()
    ax.xaxis.set_major_formatter(mdates.DateFormatter('%d-%m-%Y'))
    plt.gcf().autofmt_xdate() # Rotation
    plt.ylabel('Occupancy Count')
    plt.title(f'Manual Count vs Sensor Count for {name}')
    plt.show()

plotting within for loop, only first df is plotted. error comes after:

ConversionError: Failed to convert value(s) to axis units: 0 2022-10-27 10:30:00
1 2022-10-27 11:00:00
2 2022-10-27 11:30:00
3 2022-10-27 12:00:00
4 2022-10-27 12:30:00
...
167 2022-10-10 13:00:00
168 2022-10-10 14:00:00
169 2022-10-10 15:00:00
170 2022-10-10 16:00:00
171 NaT

Any help would be greatly appreciated.

@charlie83xt Can you please open a new issue with that example + enough (synthetic) data to reproduce the issue?