How to read a CSV file with NumPy

Delimited text is a common handoff between spreadsheets, sensors, and numerical programs. NumPy fits that boundary when every record has the same columns and the imported values need array operations rather than a dataframe layer.

The np.genfromtxt() function can take field names from the header while an explicit structured dtype keeps text, floating-point, and integer columns separate. A blank value in a floating-point field can become NaN without changing the types of the other columns.

A simple comma-delimited file should have one header row and a consistent number of fields. Files with embedded delimiters, multiline quoted fields, complex dates, or irregular records need a CSV-aware parser or a cleanup pass before their values enter the array.

1. Create a small CSV file with a header row and one blank temperature field.

   readings.csv

   sensor,temp,humidity
   A,21.5,45
   B,,52
   C,19.0,49

   The blank temp field in row B supplies a missing floating-point value for the import check.

2. Start the Python reader with the input path and structured dtype.

   csv-read.py

   from pathlib import Path
    
   import numpy as np
    
   path = Path("readings.csv")
   dtype = [("sensor", "U8"), ("temp", "f8"), ("humidity", "i8")]

   The structured dtype preserves sensor labels as Unicode, temperatures as floating-point values, and humidity readings as integers.

3. Add the NumPy loader below the dtype declaration.

   data = np.genfromtxt(
       path,
       delimiter=",",
       names=True,
       dtype=dtype,
       encoding="utf-8",
       missing_values="",
       filling_values={"temp": np.nan},
   )

   names=True consumes the header as field names, while the column-specific fill value keeps the blank temperature numeric.

4. Append the field checks after the loader call.

   print("rows:", data.shape[0])
   print("columns:", data.dtype.names)
   print("temps:", data["temp"])
   print("missing temp:", np.isnan(data["temp"][1]))
   print("average temp:", np.nanmean(data["temp"]))

   np.nanmean() excludes the missing temperature from the average while leaving the NaN visible in the imported field.

5. Run the completed script to load readings.csv into a structured NumPy array.

   $ python3 csv-read.py
   rows: 3
   columns: ('sensor', 'temp', 'humidity')
   temps: [21.5  nan 19. ]
   missing temp: True
   average temp: 20.25

   Three rows, the expected field names, a missing second temperature, and a mean of 20.25 confirm that the file crossed into typed fields without replacing the blank with a normal number.