Williams College Lecture 12 Brent Heeringa, Bill Jannen

CSV
Comma Separated Values (CSV) is a common data file format that represents data as a row of values, separated by
a delimiter, which is typically a comma. Data in spreadsheets and databases matches this format nicely, so CSV is
often used as an export format. If you work in data science, CSV is ubiquitous, so it makes sense to spend some
time learning more about the format and developing skills to manipulate data once its read into memory.
Here is some example CSV data representing financial information from Apple Computer.

Date,Open,High,Low,Close,Volume,Adj Close
2009-12-31,213.13,213.35,210.56,210.73,88102700,28.40
2009-12-30,208.83,212.00,208.31,211.64,103021100,28.52
2009-12-29,212.63,212.72,208.73,209.10,111301400,28.18
2009-12-28,211.72,213.95,209.61,211.61,161141400,28.51

header many CSV files start with an initial header row, which gives column names for the data

data data in CSVs is separated by commas, but any delimiter can be used.

Python and CSV: Readers

Suppose the the contents of the above CSV were in a file called aapl.csv. One could open that CSV and stream
through the data using the csv module and the following syntax.
1 import csv
2
3 with open(’aapl.csv’, ’r’) as fin:
4 print(list(csv.reader(fin)))

The reader object is iterable. Each row is a list of strings that were split by the delimiter, which by default is
the comma. This yields the following output.

[[’Date’, ’Open’, ’High’, ’Low’, ’Close’, ’Volume’, ’Adj Close’],

[’2009-12-31’, ’213.13’, ’213.35’, ’210.56’, ’210.73’, ’88102700’, ’28.40’],
[’2009-12-30’, ’208.83’, ’212.00’, ’208.31’, ’211.64’, ’103021100’, ’28.52’],
[’2009-12-29’, ’212.63’, ’212.72’, ’208.73’, ’209.10’, ’111301400’, ’28.18’],
[’2009-12-28’, ’211.72’, ’213.95’, ’209.61’, ’211.61’, ’161141400’, ’28.51’]]

If I wanted to find the highest stock price over the time period given by this data, I could write:
1 import csv
2 HIGHPRICECOL = 2
3
4 with open(’aapl.csv’, ’r’) as fin:
5 data = list(csv.reader(fin))
6 prices = [row[HIGHPRICECOL] for row in data[1:]]
7 print(max(prices))

This code makes several assumptions and uses some new python constructs, all of which are worth mentioning:

• all the data is held in memory at once so this is not a streaming algorithm;

• we use a list comprehension when assigning the prices variable; and

• we assume that we know the column index for price, which is 2.

Fall Semester 2016 1 CS 135: Diving into the Deluge of Data

 Williams College Lecture 12 Brent Heeringa, Bill Jannen

Let’s write this without loading all the data into memory and without knowing which column the ’High’ price
occupies:
1 import csv
2 COLNAME = ”High”
3
4 with open(’aapl.csv’, ’r’) as fin:
5 maxprice = float(’−inf’)
6 maxpricecol = None
7
8 for rownum, row in enumerate(csv.reader(fin)):
9 if rownum == 0:
10 maxpricecol = row.index(COLNAME)
11 else:
12 maxprice = max(maxprice, float(row[maxpricecol]))
13
14 print(maxprice)

This code also uses some Python constructs that are new to us. The first is float(’-inf’), which is a Python
way of specifying a value that is always smaller than any other value. The maxpricecol variable we declare and
initialize to None. The index method returns the index or position of the value ‘‘High’’ in the row. Again,
note that this algorithm is streaming, so we only consider one line at a time.

Practice
Imagine I have a file called realestate.csv that contains real estate transactions in Sacramento over 5 days.
The format of the data is

street,city,zip,state,beds,baths,sqft,type,sale date,price,lat,long.

Write a short script that finds that average sale price for every transaction in the file. You know that price occurs
at index 9, that the statistics.mean function is available, and that the data can easily fit into memory.
1 import sys
2 import csv
3 import statistics
4 PRICECOL = 9
5
6 def mean sale(filename):
7 with open(filename, ’r’) as fin:
8 rows = list(csv.reader(fin))[1:]
9 prices = [float(row[PRICECOL]) for row in rows]
10 return(statistics.mean(prices))
11
12 if name == ’ main ’:
13 print(mean sale(sys.argv[1]))

Now imagine writing a function that returned the mean sale price of houses over 2000 square feet. Square
footage is given by the column at index 6.
1 import csv
2 import sys
3 import statistics

Fall Semester 2016 2 CS 135: Diving into the Deluge of Data

 Williams College Lecture 12 Brent Heeringa, Bill Jannen

4
5 PRICECOL = 9
6 SQFTCOL = 6
7 SQFTMIN = 2000
8
9 def mean sale high(filename):
10 with open(filename, ’r’) as fin:
11 rows = list(csv.reader(fin))[1:]
12 prices = []
13 for row in rows:
14 if int(row[SQFTCOL]) > SQFTMIN:
15 prices.append(float(row[PRICECOL]))
16 return(statistics.mean(prices))
17
18 if name == ’ main ’:
19 print(mean sale(sys.argv[1]))

You can also use an if statement in the list comprehension for some truly beautiful code.
1 def mean sale high2(filename):
2 with open(filename, ’r’) as fin:
3 rows = list(csv.reader(fin))[1:]
4 prices = [float(row[9]) for row in rows if int(row[6]) > 2000]
5 print(statistics.mean(prices))

CSV Data in Strings

Suppose that the CSV data, however, is in a string data, instead of a file. In this case, one would use the
io.StringIO type to wrap the string inside something that behaves like a file object. You can think of this
as buffering the string.
1 import csv
2 import io
3
4 data = ’purple,cow,moo\nhappy,moose,grunt’
5 reader = csv.reader(io.StringIO(data))
6 for row in reader:
7 print(”∗”.join(row))

Reader Options
There are many options when creating a CSV reader. Here are some, with definitions coming directly from the API1 :

delimiter A one-character string used to separate fields. It defaults to ’,’.

escapechar On reading, the escapechar removes any special meaning from the following character. It de-
faults to None, which disables escaping.
lineterminator The string used to terminate lines produced by the writer. It defaults to ’\r\n’. Note The
reader is hard-coded to recognize either ’\r’ or ’\n’ as end-of-line, and ignores line terminator. This
behavior may change in the future.
1
https://docs.python.org/3.4/library/csv.html#csv-fmt-params

Fall Semester 2016 3 CS 135: Diving into the Deluge of Data

 Williams College Lecture 12 Brent Heeringa, Bill Jannen

As an extreme example, suppose we wanted to represent a bunch of data that was just commas. One could use a
different delimiter

,|,,|,
,,|,|,,

and use csv.reader(filename.csv, delimiter="|") to create the correct reader. We could also escape
the commas

\,,\,\,,\,
\,\,,\,,\,\,

and use csv.reader(filename.csv, escapechar="\\") to create the correct reader. Notice that we
need to escape the backslash inside the character string.

Writers
CSV Writer objects accept any object that has a write method (file objects, StringIO objects, etc.) and formats
CSV data using the writerow or writerows method. Here’s an example. Suppose that data is a list of NESCAC
school information.

data = [[’Williams’, ’Ephs’, ’Purple Cows’],

[’Middlebury’, ’Panthers’, ’Panther’]]

To write this to the file called nescac.csv we would use the following code
1 import csv
2 with open(’nescac.csv’, ’w’, newline=’’) as csvfile:
3 writer = csv.writer(csvfile, delimiter=’,’)
4 writer.writerow([’School’, ’Nickname’, ’Mascot’])
5 writer.writerows(data)

Practice
Suppose you had a list of constellations and their galactic coordinates (right ascension and declination) in CSV
format.

constellation, right ascension, declination

Sagittarius,19,-25
Taurus, 4.9, 19
Perseus, 3, 45

Write a function that takes a filename file in CSV format and returns a list of constellations. Suppose that you
know one of the headers is labelled constellation, but not which one. Suppose further that you can easily fit
all the data in memory.
1 with open(file, newline=’’) as fp:
2 data = [row for row in csv.reader(file)]
3 col = data[0].index(’constellation’)
4 return [row[col] for row in data[1:]]

Fall Semester 2016 4 CS 135: Diving into the Deluge of Data