python - Matplotlib dates YearLocator with odd intervals

Question

Right now I have a working bit of code that formats to my specification when my time series begin at the start of a decade (i.e. 1990, 2000, 2010,etc.), but I don't know how to adapt my code to have the correct formatting when my time series begins at a year that isn't even (i.e. 1993).

import pandas as pd
import matplotlib.pyplot as plt
from matplotlib import dates

def format_xaxis(fig):

     years = dates.YearLocator(10,month=1,day=1)
     years1=dates.YearLocator(2,month=1,day=1)
     dfmt = dates.DateFormatter('%Y')
     dfmt1 = dates.DateFormatter('%y')

     [i.xaxis.set_major_locator(years) for i in fig.axes]
     [i.xaxis.set_minor_locator(years1) for i in fig.axes]
     [i.xaxis.set_major_formatter(dfmt) for i in fig.axes]
     [i.xaxis.set_minor_formatter(dfmt1) for i in fig.axes]
     [i.get_xaxis().set_tick_params(which='major', pad=15) for i in fig.axes]

     for t in fig.axes:
         for tick in t.xaxis.get_major_ticks():
             tick.label1.set_horizontalalignment('center')
         for label in t.get_xmajorticklabels() :
             label.set_rotation(0)
             label.set_weight('bold')
         for label in t.xaxis.get_minorticklabels():
             label.set_fontsize('small')
         for label in t.xaxis.get_minorticklabels()[::5]:
             label.set_visible(False)


df = pd.DataFrame.from_dict({'Y': {0: 0.15,  1: 0.18,  2: 0.23,  3: 0.15,  4: 0.15,  5: 0.15,  6: 0.17,  7: 0.175,  8: 0.212,  9: 0.184,  10: 0.18,  11: 0.18,  12: 0.21,  13: 0.139,  14: 0.15,  15: 0.128,  16: 0.126,  17: 0.1,  18: 0.11,  19: 0.183,  20: 0.14,  21: 0.12,  22: 0.155,  23: 0.245,  24: 0.248,  25: 0.262,  26: 0.17,  27: 0.143,  28: 0.13,  29: 0.102,  30: 0.258,  31: 0.293,  32: 0.196,  33: 0.21,  34: 0.14,  35: 0.17}, 
                             'Date': {0: '1990-06-10 00:00:00',  1: '1991-07-26 00:00:00',  2: '1992-10-15 00:00:00',  3: '1993-10-08 00:00:00',  4: '1994-04-07 00:00:00',  5: '1994-11-20 00:00:00',  6: '1995-04-24 00:00:00',  7: '1996-02-13 00:00:00',  8: '1996-04-15 00:00:00',  9: '1996-09-12 00:00:00',  10: '1997-02-13 00:00:00',  11: '1997-04-20 00:00:00',  12: '1997-08-23 00:00:00',  13: '1997-11-06 00:00:00',  14: '1998-04-15 00:00:00',  15: '1999-05-04 00:00:00',  16: '2000-03-17 00:00:00',  17: '2000-06-01 00:00:00',  18: '2001-10-05 00:00:00',  19: '2002-09-20 00:00:00',  20: '2003-04-25 00:00:00',  21: '2003-09-20 00:00:00',  22: '2005-05-07 00:00:00',  23: '2006-10-07 00:00:00',  24: '2007-10-13 00:00:00',  25: '2008-02-02 00:00:00',  26: '2008-03-28 00:00:00',  27: '2008-10-10 00:00:00',  28: '2009-10-10 00:00:00',  29: '2011-10-05 00:00:00',  30: '2012-10-03 00:00:00',  31: '2013-09-21 00:00:00',  32: '2014-09-23 00:00:00',  33: '2015-09-22 00:00:00',  34: '2016-10-01 00:00:00',  35: '2017-09-29 00:00:00'}})

df['Date'] = pd.to_datetime(df['Date'], format='%Y-%m-%d %H:%M:%S')

fig, ax = plt.subplots()

df.plot('Date','Y',ax=ax,marker='x',ls='-')
ax.set_xlim(pd.datetime(1990, 1, 1), pd.datetime(2018, 1, 1))

format_xaxis(fig)

Which produces a plot that looks like:

How would I recreate the above plot with a time series that begins on 1993? I would still like the minor tick labels to be set every two years (i.e. 95,97,99,01,....). Is it possible to use matplotlib.dates.YearLocator for format dates when a time series plot begins on an odd year?

Answer 1

You may subclass YearLocator to have your custom OffsetYearLocator.

Version for matplotlib <= 3.0

from matplotlib import dates

class OffsetYearLocator(dates.YearLocator):
    def __init__(self, *args, **kwargs):
        self.offset = kwargs.pop("offset", 0)
        dates.YearLocator.__init__(self,*args, **kwargs)
    def tick_values(self, vmin, vmax):
        ymin = self.base.le(vmin.year)-self.offset
        ymax = self.base.ge(vmax.year)+(self.base._base-self.offset)
        ticks = [vmin.replace(year=ymin, **self.replaced)]
        while True:
            dt = ticks[-1]
            if dt.year >= ymax:
                return dates.date2num(ticks)
            year = dt.year + self.base.get_base()
            ticks.append(dt.replace(year=year, **self.replaced))

Version for matplotlib >= 3.1

from matplotlib import dates

class OffsetYearLocator(dates.YearLocator):
    def __init__(self, *args, **kwargs):
        self.offset = kwargs.pop("offset", 0)
        dates.YearLocator.__init__(self,*args, **kwargs)
    def tick_values(self, vmin, vmax):
        ymin = self.base.le(vmin.year) * self.base.step - self.offset
        ymax = self.base.ge(vmax.year) * self.base.step + (self.base.step-self.offset)
        ticks = [vmin.replace(year=ymin, **self.replaced)]
        while True:
            dt = ticks[-1]
            if dt.year >= ymax:
                return dates.date2num(ticks)
            year = dt.year + self.base.step
            ticks.append(dt.replace(year=year, **self.replaced))

This can handle an additional argument offset, which is substracted from the year. In this case, one would keep the base as 2 (every two years), but use an offset of 1.

years1 = OffsetYearLocator(2, month=1, day=1, offset=1)

Complete example (using the version for matplotlib >= 3.1):

import pandas as pd
import matplotlib.pyplot as plt
from matplotlib import dates

class OffsetYearLocator(dates.YearLocator):
    def __init__(self, *args, **kwargs):
        self.offset = kwargs.pop("offset", 0)
        dates.YearLocator.__init__(self,*args, **kwargs)
    def tick_values(self, vmin, vmax):
        ymin = self.base.le(vmin.year) * self.base.step - self.offset
        ymax = self.base.ge(vmax.year) * self.base.step + (self.base.step-self.offset)
        ticks = [vmin.replace(year=ymin, **self.replaced)]
        while True:
            dt = ticks[-1]
            if dt.year >= ymax:
                return dates.date2num(ticks)
            year = dt.year + self.base.step
            ticks.append(dt.replace(year=year, **self.replaced))

def format_xaxis(ax):

    years = dates.YearLocator(10,month=1,day=1)
    years1=OffsetYearLocator(2,month=1,day=1, offset=1)
    dfmt = dates.DateFormatter('%Y')
    dfmt1 = dates.DateFormatter('%y')

    ax.xaxis.set_major_locator(years)
    ax.xaxis.set_minor_locator(years1)
    ax.xaxis.set_major_formatter(dfmt)
    ax.xaxis.set_minor_formatter(dfmt1)
    ax.get_xaxis().set_tick_params(which='major', pad=15)

    plt.setp(ax.get_xmajorticklabels(), rotation=0, weight="bold", ha="center")


df = pd.DataFrame.from_dict({'YData': {0: 0.15,  1: 0.18,  2: 0.23,  3: 0.15,  4: 0.15,  5: 0.15,  6: 0.17,  7: 0.175,  8: 0.212,  9: 0.184,  10: 0.18,  11: 0.18,  12: 0.21,  13: 0.139,  14: 0.15,  15: 0.128,  16: 0.126,  17: 0.1,  18: 0.11,  19: 0.183,  20: 0.14,  21: 0.12,  22: 0.155,  23: 0.245,  24: 0.248,  25: 0.262,  26: 0.17,  27: 0.143,  28: 0.13,  29: 0.102,  30: 0.258,  31: 0.293,  32: 0.196,  33: 0.21,  34: 0.14,  35: 0.17}, 
                             'Date': {0: '1990-06-10 00:00:00',  1: '1991-07-26 00:00:00',  2: '1992-10-15 00:00:00',  3: '1993-10-08 00:00:00',  4: '1994-04-07 00:00:00',  5: '1994-11-20 00:00:00',  6: '1995-04-24 00:00:00',  7: '1996-02-13 00:00:00',  8: '1996-04-15 00:00:00',  9: '1996-09-12 00:00:00',  10: '1997-02-13 00:00:00',  11: '1997-04-20 00:00:00',  12: '1997-08-23 00:00:00',  13: '1997-11-06 00:00:00',  14: '1998-04-15 00:00:00',  15: '1999-05-04 00:00:00',  16: '2000-03-17 00:00:00',  17: '2000-06-01 00:00:00',  18: '2001-10-05 00:00:00',  19: '2002-09-20 00:00:00',  20: '2003-04-25 00:00:00',  21: '2003-09-20 00:00:00',  22: '2005-05-07 00:00:00',  23: '2006-10-07 00:00:00',  24: '2007-10-13 00:00:00',  25: '2008-02-02 00:00:00',  26: '2008-03-28 00:00:00',  27: '2008-10-10 00:00:00',  28: '2009-10-10 00:00:00',  29: '2011-10-05 00:00:00',  30: '2012-10-03 00:00:00',  31: '2013-09-21 00:00:00',  32: '2014-09-23 00:00:00',  33: '2015-09-22 00:00:00',  34: '2016-10-01 00:00:00',  35: '2017-09-29 00:00:00'}})

df['Date'] = pd.to_datetime(df['Date'], format='%Y-%m-%d %H:%M:%S')

fig, ax = plt.subplots()

ax.plot('Date','YData', data=df, marker='x',ls='-')
ax.set_xlim(pd.datetime(1990,1,1), pd.datetime(2018,1,1))

format_xaxis(ax)

plt.show()