Timekeeping in Python, R¶

Clocks and timekeeping on computers are not trivial.¶

Remember Y2k? All of that was caused by the representation of 2-digit years in computer systems.
For many years, all timekeeping on computer systems was referenced to January 1, 1970 because no computer data pre-existed that. This is known as the POSIX time.

Python has multiple timekeeping modules including base python modules date, time, datetime. We'll focus on datetime.

Note: Numpy and Pandas also have their own timekeeping, but both are compatible with the base python package of datetime.

In [ ]:

# PYTHON
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import time as time
import datetime as dt

In [ ]:

# R
#install.packages("lubridate")
library(lubridate)

Da = ymd("2019-12-13")
Db <- ymd_hms("2000-12-13 13:55:59")

print(Da)

In [ ]:

# A Datetime object can incorporate Year, Month, Day, Hour, Minute, Second, Millisecond.
# At a minimum, year, month, day must be specified

Da = dt.datetime(
    year=2019,\
    month=12,\
    day=13,\
    second=59
)
print(Da)


# Specify everything in order of Year, Month, Day, Hour, Minute, Second
Db = dt.datetime(2000,12,13,13,55,59)
print(Db)


# Specify less than the minimum
Dc = dt.datetime(2020,12,1)
print(Dc)

Datetime objects support basic arithmetic:

In [ ]:

# PYTHON,
# Subtracting 
Dg = Da-Db
print(Dg)
# Dg is a timedelta object
Dg?
Dg.total_seconds()


# Datetime and timedelta objects can be added and subtracted
print( dt.datetime(1990,10,31)+dt.timedelta(days=33) )
print( dt.datetime(1990,10,31)-Dg )

# This is not permitted, because Db and Da represent absolute time since a reference.
Df = Db+Da
print(Df)
Df?
Df.total_seconds()

In [ ]:

Dg?

In [ ]:

# Datetime provides the current time stamp using now()
t0 = dt.datetime.now()
print(t0)

Time references:¶

Datetime uses the proleptic Gregorian calendar as a reference. Year 1, is the reference year. Year 2020 is Year 2020.
This differs from e.g. the POSIX calendar, which uses 1970 as the reference year.
Unless specified, the datetime object is naive, meaning it doesn't know its time relative to geography.

In [ ]:

# R
# Datetime provides the current time stamp using now()
t0 <- ymd_hms(Sys.time()) 
print(t0)
t1 <- as.integer(t0) # This creates seconds from the start of the POSIX calendar.
print(t1)

In [ ]:

# Express t0 to total number of elapsed days since January 1, of year 1.
t0 = dt.datetime.now()
t1 = t0.toordinal() # This creates days from the proleptic Gregorian calendar.
# t1 is an integer.
print(t1)

# Recover timestamp, but note that hrs,mins,seconds have been lost
print(dt.datetime.fromordinal(t1))

# Express t0 as total seconds elapsed since January 1, of year 1 (the reference year).
t0.timestamp()   # This value is a float, and its the number of seconds from the start of the POSIX calendar.

In [ ]:

# Include notes on tzinfo()

# Can add and subtract datetime objects.  These are stored as timedelta()
dt.datetime.now().timestamp()

In [ ]:

# PYTHON
print(t0.tzinfo==None)
# pytz gives the time zone information
import pytz
timezone = pytz.timezone("America/New_York")
ttz = timezone.localize(t0)
print(ttz)  #this prints offset from UTC.

In [ ]:

# R
# Check if t0 has a time zone
print(is.null(attr(t0, "tzone")))

# Set time zone to America/New_York
ttz <- with_tz(t0, "America/New_York")

# Print the datetime with time zone information
print(ttz)

In [ ]:

# PYTHON
# Datetime objects can be converted to strings.
print(t0)
T = t0.strftime('%m/%d/%y,%H:%M:%S')
print(T)

In [ ]:

# R
# Format the datetime as a string
print(t0)
T <- format(t0, "%m/%d/%y,%H:%M:%S")

# Print the formatted string
print(t0)

Recording datetime to make a timeseries¶

Dataframes can incorporate the datetime object directly into its dataframe. In Python, Pandas is one library that was built for working with dataframes. In R, base R creates dataframes by default.

In [ ]:

# Python
dd = pd.DataFrame(columns=('time','iterator')) # Create a pandas dataframe with 2 columns.


for i in range(0,10):
    t = dt.datetime.now()
    dd = dd.append({'time': t,'iterator': i}, ignore_index=True)
    #dd = dd.append({}, ignore_index=True)
    time.sleep(0.5)
print(dd)
dd.head()

In [ ]:

# Python
# Importantly, the datetime vector can be plotted in matplotlib.
plt.figure()
plt.plot(dd['time'],dd['iterator'],'.')
plt.xlabel('time'); plt.ylabel('iterator')
plt.show()

In [ ]:

# R
# Create an empty data frame with columns 'time' and 'iterator'
dd <- data.frame(time = character(), iterator = numeric())

# Loop 10 times
for (i in 1:10) {
  # Get current datetime
  t <- ymd_hms(Sys.time())
  
  # Append a new row with current time and iterator value
  dd <- rbind(dd, list(time = t, iterator = i))
  
  # Sleep for 0.5 seconds
  Sys.sleep(1)
}

# Print the entire data frame
print(dd)

# Print the first 5 rows
head(dd)

In [ ]:

# R
# NOTE:  R doesn't have datetime as a data type, so it is harder to graph date and time on one axis.  
# This example graphs them in numeric seconds
library(ggplot2)
# Create a scatter plot
ggplot(dd, aes(x = time, y = iterator)) +
  geom_point(color = "blue") +  # Plot points with blue color
  labs(x = "Time", y = "Iterator") +  # Set axis labels
  theme_minimal()  # Apply a clean theme