[PYTHON] Change the Y-axis scale of Matplotlib to exponential notation (10 Nth power notation)

ex1) Draw a graph

test1.py

import sqlalchemy
import pandas as pd
import matplotlib.pyplot as plt

#SQLAlchemy initialization
CONNECT_INFO = 'mssql+pyodbc://hogehoge'
engine = sqlalchemy.create_engine(CONNECT_INFO, encoding='utf-8')

#Variable setting
qq = 1
ym = 201607

#DB connection, stored execution, pandas data frame creation
query = 'EXEC dbo.sp_rtrv_hogehoge @q = {0},@prd = {1}'.format(qq,ym )
df = pd.read_sql_query(query, engine  ,index_col =['t'])

#Graph drawing
ax = df.plot(  color=('b','r') , alpha=0.6  )
plt.title('hogehoge_title')
plt.grid(which='major')
#ax = plt.gca()

#Save file
fname ='test'+ str(ym) +'.png'
plt.savefig(fname) 
plt.close()  

-Plot () of pandas (df) is a wrapper method of plt.plot () of matplotlib. -In the above, the SQLSevrer stored procedure is executed (in SQL Server, the stored procedure is executed by EXEC). -Stored arguments (qq and ym) are specified in python variable embedded notation. -You can get the graph currently being edited with ʻax = plt.gca () `

-In the above, no Y-axis scale is applied, so for example, in the case of 1 million, it is written as "1000000" as it is.

Y-axis scale ↓

ex2) Change the Y-axis scale to scientific notation

test1.py

import sqlalchemy
import pandas as pd
import matplotlib.pyplot as plt

#(Omission)

#Graph drawing
ax = df.plot(  color=('b','r') , alpha=0.6  )
#ax = plt.gca()
ax.ticklabel_format(style="sci",  axis="y",scilimits=(0,0))

Y-axis scale ↓

・ It became a scale like "Data Scientist" (laughs) However, with the notation of le6, there is a risk that the boss of the liberal arts Osan, who is weak in mecha, will become a boon (white eyes), so the next ex3 was fine-tuned.

ex3) Change the Y-axis scale to a power of 10 notation

test1.py

import sqlalchemy
import pandas as pd
import matplotlib.pyplot as plt
from matplotlib.ticker import ScalarFormatter

#(Omission)

#Graph drawing
ax = df.plot(  color=('b','r') , alpha=0.6  )
#ax = plt.gca()
ax.yaxis.set_major_formatter(ScalarFormatter(useMathText=True))
ax.ticklabel_format(style="sci",  axis="y",scilimits=(0,0))

-If you set the property of ScalarFormatter to ʻuseMathText = True`, it becomes a power notation of 10. ・ One million is $ 1 × 10 ^ 6 $ (** 10 to the 6th power **, that is, 6 zeros)

Y-axis scale ↓

ex4) Change the Y-axis scale from 1 × 10 ^ 6 to 1 × 10 ^ 4.

test4.py

import sqlalchemy
import pandas as pd
import matplotlib.pyplot as plt
from matplotlib.ticker import ScalarFormatter

#Class settings * Inherit ScalarFormatter
class FixedOrderFormatter(ScalarFormatter):
    def __init__(self, order_of_mag=0, useOffset=True, useMathText=True):
        self._order_of_mag = order_of_mag
        ScalarFormatter.__init__(self, useOffset=useOffset, 
                                 useMathText=useMathText)
    def _set_orderOfMagnitude(self, range):
        self.orderOfMagnitude = self._order_of_mag

#(Omission)

#Graph drawing
ax = df.plot(  color=('b','r') , alpha=0.6  )
#ax = plt.gca()
ax.yaxis.set_major_formatter(FixedOrderFormatter(4 ,useMathText=True))
#ax.yaxis.set_major_formatter(ScalarFormatter(useMathText=True))
ax.ticklabel_format(style="sci",  axis="y",scilimits=(0,0))

・ Depending on the data, you may want to use 10,000 units ($ 1 x 10 ^ 4 ) instead of 1 million notation ( 1 x 10 ^ 6 $)! ・ When I searched and searched various things such as matplotlib's github and stackOverflow, it seems that ScalarFormatter has an orderOfMagnitude, and here it controls the N part of 10 Nth power. -In the above, the class that inherits ScalarFormatter is set, and the state of ʻuseMathText = True` is kept as it is, and the orderOfMagnitude of ScalarFormatter is specified and changed to 10 4 notation.

Y-axis scale ↓

・ 10,000 is $ 1 × 10 ^ 4 $ (** 10 to the 4th power **, that is, 4 zeros), so 1 million is $ 100 × 10 ^ 4 $

Reference URL

・ [Matplotlib] I want to use the scale of the axis as exponential notation (via. · Matplotlib: format axis offset-values to whole numbers or specific numbers (Via.stackOverflow) ・ Github: matplotlib / lib / matplotlib / ticker.py