How can I use matplotlib ticklabel_format to not use scientific notation on y axis labels - matplotlib

Question

I'm creating a plot (in a colab worksheet) and want the y tick labels to not use scientific notation. The ticklabel_format doesn't make any difference to the final graph. The y axis labels are still shown as 10^3 instead of 1000. How do I format the y tick labels to not use scientific notation?
Here is my code
import matplotlib.pyplot as plt
plt.ticklabel_format(style='plain', axis='y')
plt.plot(Cd_rank,Cd_raw,linewidth=4)
plt.plot(Cd_rank,Cd_sed,linewidth=4)
plt.plot(Cd_rank,Cd_filter,linewidth=4)
plt.plot([0,1],[0.3,0.3],linewidth=4)
plt.plot([0,1],[5,5],linewidth=4)
plt.ylabel('Turbidez (UTN)')
plt.xlabel('Datos ordenados')
plt.yscale('log')
plt.legend(['Agua cruda','Decantada','Filtrada','Norma EPA','Norma ENACAL'])

Answer 1

The ScalarFormatter shows the tick labels in a default format. Note that depending on your concrete situation, matplotlib still might be using scientific notation:
When the numbers are too high (default this is about 4 digits). set_powerlimits((n, m)) can be used to change the limits.
In case the numbers are very close together, matplotlib describes the range using an offset. That offset is placed at the top of the axis. This can be suppressed with the useOffset=None parameter of the formatter.
In some cases with a logarithmic scale, there are very few major ticks. Then also some (but not all) minor ticks get a label. Also for these, the formatter could be changed. A problem can be that a simple ScalarFormatter will set too many labels. Either suppress all these minor labels using a NullFormatter or you'll need a very custom formatter that returns empty strings for the minor tick labels that need to be suppressed.
A simple example:
from matplotlib import pyplot as plt
from matplotlib import ticker
import numpy as np
N = 50
Cd_rank = np.linspace(0, 100, N)
Cd_raw = np.random.normal(1, 20, N).cumsum() + 100
plt.plot(Cd_rank, Cd_raw, linewidth=4)
plt.plot([0, 1], [0.3, 0.3], linewidth=4)
plt.plot([0, 1], [5, 5], linewidth=4)
plt.yscale('log')
plt.gca().yaxis.set_major_formatter(ticker.ScalarFormatter())
plt.gca().yaxis.set_minor_formatter(ticker.NullFormatter())
plt.show()
And here is a more complicated example, with both minor (green) and major (red) ticks.
from matplotlib import pyplot as plt
from matplotlib import ticker
import numpy as np
N = 50
Cd_rank = np.linspace(0, 100, N)
Cd_raw = np.random.normal(10, 5, N).cumsum() + 80
plt.plot(Cd_rank, Cd_raw, linewidth=4)
plt.yscale('log')
mticker = ticker.ScalarFormatter(useOffset=False)
mticker.set_powerlimits((-6, 6))
ax = plt.gca()
ax.yaxis.set_major_formatter(mticker)
ax.yaxis.set_minor_formatter(mticker)
ax.tick_params(axis='y', which='major', colors='crimson')
ax.tick_params(axis='y', which='minor', colors='seagreen')
plt.show()
PS: When the ticks involve both powers of 10 larger than 1 and smaller than 1 (so, e.g. 100, 10, 1, 0.1, 0.01) the ScalarFormatter doesn't display the numbers smaller than 1 well (it displays 0.1 and 0.01 as 0). In that case, the StrMethodFormatter can be used instead:
plt.gca().yaxis.set_major_formatter(ticker.StrMethodFormatter("{x}"))

Answer 2

Here is code that turns off scientific notation and handles numbers that are smaller than 1 correctly. Thanks to #Johanc for this code.
from matplotlib import pyplot as plt
from matplotlib import ticker
import numpy as np
N = 50
x = np.linspace(0,1,N)
y = np.logspace(-3, 2, N)
plt.plot(x, y, linewidth=4)
plt.yscale('log')
plt.ylim(bottom=0.001,top=100)
plt.gca().yaxis.set_major_formatter(ticker.ScalarFormatter())
plt.gca().yaxis.set_major_formatter(ticker.StrMethodFormatter("{x}"))
plt.show()```