Reading CSV Files (read_csv)

Real cricket data never starts life inside your Python code. The IPL ball-by-ball dataset, the Orange Cap table, a season's match results — they all arrive as files, and the most common file of all is the CSV (Comma-Separated Values). A CSV is just plain text: the first line usually holds the column names, and every following line is one row, with values separated by commas.

player,team,matches,runs,average
Virat Kohli,RCB,15,741,61.75
Ruturaj Gaikwad,CSK,14,583,53.0

pd.read_csv() reads that text, splits every line on the separator, guesses a data type for each column and hands you back a ready-to-use DataFrame. It is the single most-used function in all of pandas — almost every analysis begins with it.

Definition

pd.read_csv() is the pandas function that reads delimited text data (a CSV file, a URL, or any file-like object) and returns it as a DataFrame, automatically detecting column names from the header row and inferring a data type for every column.

• Delimited text — values separated by a fixed character. Usually a comma, but semicolons (;), tabs and pipes (|) are also common. The sep parameter tells pandas which one.
• File-like object — anything with a .read() method. That is why we can hand it an io.StringIO (a string pretending to be a file) — perfect for the browser playground, which has no real files.
• Header row — by default line 1 becomes the column names (header=0).
• Infers a data type — a column of whole numbers becomes int64, decimals become float64, and text becomes the new str dtype in pandas 3.

Syntax

pd.read_csv(filepath_or_buffer, sep=",", header="infer", names=None,
            index_col=None, usecols=None, dtype=None, parse_dates=None,
            na_values=None, nrows=None, skiprows=None, encoding=None)

This topic covers the "shape of the file" parameters. The next topic covers the rest.

Reading a real file — three ways to write a Windows path

In Munna's notebooks the file sits on the Desktop. A Windows path contains backslashes, and in Python \ starts an escape sequence (\n, \t…), so write the path in one of these three safe forms:

df = pd.read_csv(r"C:\Users\Admin\Desktop\data\ipl_matches.csv")    # 1. raw string (r"...")
df = pd.read_csv("C:\\Users\\Admin\\Desktop\\data\\ipl_matches.csv") # 2. double backslashes
df = pd.read_csv("C:/Users/Admin/Desktop/data/ipl_matches.csv")      # 3. forward slashes

All three work. A URL works too — pd.read_csv("https://.../matches.csv") downloads and reads the file in one step (for GitHub, use the raw file link).

Simulating a file with io.StringIO

The browser playground has no Desktop and no internet, so in these notes we wrap CSV text in io.StringIO. To read_csv, it looks exactly like an open file:

import pandas as pd
import io

csv_text = """player,team,matches,runs,average
Virat Kohli,RCB,15,741,61.75
Ruturaj Gaikwad,CSK,14,583,53.0
Riyan Parag,RR,16,573,52.09
Travis Head,SRH,15,567,40.5
Sanju Samson,RR,16,531,48.27
"""
df = pd.read_csv(io.StringIO(csv_text))
print(df)
print(df.dtypes)

            player team  matches  runs  average
0      Virat Kohli  RCB       15   741    61.75
1  Ruturaj Gaikwad  CSK       14   583    53.00
2      Riyan Parag   RR       16   573    52.09
3      Travis Head  SRH       15   567    40.50
4     Sanju Samson   RR       16   531    48.27
player         str
team           str
matches      int64
runs         int64
average    float64
dtype: object

How it works: pandas read the first line as column names, created a default RangeIndex (0, 1, 2…) for the rows, and guessed the types: matches and runs hold whole numbers → int64; average has decimals → float64; names and teams are text → str.

Note (pandas 3): text columns now show the dtype str. In pandas 2 (and in most older tutorials) the same columns showed object. Also, type(df) now prints <class 'pandas.DataFrame'> instead of pandas.core.frame.DataFrame.

Example — a different separator (sep)

Many European and Excel-exported files use a semicolon. If you forget sep, pandas treats every whole line as one value:

semi = """player;team;runs
Virat Kohli;RCB;741
Riyan Parag;RR;573
Travis Head;SRH;567
"""
print(pd.read_csv(io.StringIO(semi)))             # wrong: default sep=","
print(pd.read_csv(io.StringIO(semi), sep=";"))    # right

      player;team;runs
0  Virat Kohli;RCB;741
1   Riyan Parag;RR;573
2  Travis Head;SRH;567
        player team  runs
0  Virat Kohli  RCB   741
1  Riyan Parag   RR   573
2  Travis Head  SRH   567

How it works: the first result has shape (3, 1) — a single column called player;team;runs. That "one giant column" is the classic sign of a wrong separator. For tab-separated files use sep="\t" (or the shortcut pd.read_table(), whose default separator is a tab).

Example — files without a header (header=None, names)

Some exported files contain only data rows. By default pandas would wrongly promote the first player to column names:

nohead = """Virat Kohli,RCB,741
Riyan Parag,RR,573
Travis Head,SRH,567
"""
print(pd.read_csv(io.StringIO(nohead)))                       # first row eaten as header!
print(pd.read_csv(io.StringIO(nohead), header=None))          # numbered columns 0, 1, 2
print(pd.read_csv(io.StringIO(nohead), header=None,
                  names=["player", "team", "runs"]))          # our own names

   Virat Kohli  RCB  741
0  Riyan Parag   RR  573
1  Travis Head  SRH  567
             0    1    2
0  Virat Kohli  RCB  741
1  Riyan Parag   RR  573
2  Travis Head  SRH  567
        player team  runs
0  Virat Kohli  RCB   741
1  Riyan Parag   RR  573
2  Travis Head  SRH  567

How it works: header=None says "there is no header line — every line is data", so pandas numbers the columns 0, 1, 2. Adding names=[...] gives them proper labels. You can also use names together with header=0 to replace an existing header with your own names.

Example — junk lines at the top (header=1 / skiprows)

Reports often start with a title line before the real header:

title = """IPL 2024 Orange Cap Table
player,team,runs
Virat Kohli,RCB,741
Riyan Parag,RR,573
"""
print(pd.read_csv(io.StringIO(title), header=1))     # header is on line 1 (0-based)
print(pd.read_csv(io.StringIO(title), skiprows=1))   # or: skip 1 line, then read normally

        player team  runs
0  Virat Kohli  RCB   741
1  Riyan Parag   RR   573
        player team  runs
0  Virat Kohli  RCB   741
1  Riyan Parag   RR   573

How it works: both give the same result. header=1 means "line number 1 (counting from 0) is the header and everything above it is thrown away". skiprows=1 removes the first line first, then reads the rest with the default header.

The family of read functions

Common mistakes

• ❌ pd.read_csv("C:\Users\new_data\ipl.csv") written with single backslashes — \n becomes a newline and the path breaks. ✅ Use r"...", double backslashes, or forward slashes.
• ❌ Getting one giant column like player;team;runs. ✅ The separator is wrong — pass sep=";" (or "\t", "|").
• ❌ The first data row has become your column names. ✅ The file has no header — use header=None and names=[...].
• ❌ pd.read_csv(csv_text) with the text itself. ✅ pandas treats a plain string as a file path; wrap text in io.StringIO(csv_text).

Interview tip

Q: Your CSV loads as a DataFrame with only one column. What happened? A: The file uses a different delimiter (often ; or a tab). pandas split on commas, found none, and kept each whole line as a single value. Fix it with sep=";" or sep="\t". Checking df.shape right after loading catches this immediately.