Getting Help

NoteQuestions
  • How do I find out what a function does?
  • How do I search for answers when I am stuck?
  • How do I read package documentation?
NoteObjectives
  • Use help() and Jupyter’s inline help to read function documentation.
  • Find the official documentation for NumPy, SciPy, and Matplotlib.
  • Form effective search queries for Python errors and tasks.
  • Adapt code examples found online to your own problem.

help() and Jupyter inline help

Every built-in function and most library functions include built-in documentation. Call help() and pass the function name without parentheses to display it in the terminal or notebook output:

help(round)
Help on built-in function round in module builtins:

round(number, ndigits=None)
    Round a number to a given precision in decimal digits.
    
    The return value is an integer if ndigits is omitted or None.  Otherwise
    the return value has the same type as the number.  ndigits may be negative.

In the Jupyter Notebook there are two faster shortcuts:

  • Place the cursor inside a function’s parentheses and press Shift+Tab for a compact pop-up summary.
  • Type the function name followed by ? and run the cell for the full docstring.

These shortcuts work for any imported function, including NumPy and SciPy functions:

import numpy
numpy.linspace?

Official package documentation

For anything beyond a quick signature check, the official documentation sites are far richer than help() output. They include conceptual guides, worked examples, and cross-references between related functions.

When you know the function name, search the reference page directly. When you know what you want to do but not which function does it, start with the gallery (for plots) or the module overview pages.

Searching online effectively

A well-formed search query usually gets a useful answer within the first few results. A few habits help:

Include the library name. "numpy sort 2d array by column" returns far more relevant results than "sort array by column".

Paste the error type, not the full traceback. The key text is usually the last line — ValueError: operands could not be broadcast together — combined with the function you were calling. Search for "numpy ValueError operands could not be broadcast" rather than copying hundreds of lines of traceback.

Use exact Python syntax in quotes. Searching for python "numpy.loadtxt" skiprows narrows results to pages that literally discuss that argument.

Stack Overflow (stackoverflow.com) is the most reliably useful community resource. Most common problems have already been asked and answered — check the accepted answer (green tick) and the highest-voted answers, which sometimes improve on it. The tags [numpy], [scipy], and [matplotlib] narrow searches within Python questions.

Adapting examples to your problem

Documentation examples and Stack Overflow answers are written for a generic audience — they rarely match your specific data or variable names exactly. When adapting code you find:

  1. Run the example as-is first. Confirm you understand what it does before modifying it.
  2. Change one thing at a time. Swap in your variable name, run, check, then make the next change.
  3. Find documentation for any keyword arguments you do not recognise. A line like numpy.loadtxt(f, delimiter=',', usecols=(0,2)) has three arguments: look up usecols if you have not seen it before.
  4. Check the output shape and type. Many errors come from mismatches between what a function returns and what the next line expects. print(type(result), result.shape) is a quick sanity check.
CautionChallenge

Reading Documentation

Use help(sorted) or the Python documentation to answer:

  1. What arguments does sorted accept besides the iterable?
  2. How would you sort a list of strings in reverse alphabetical order?

sorted accepts key (a function applied to each element before comparison) and reverse (a boolean, default False):

words = ['banana', 'apple', 'cherry']
print(sorted(words, reverse=True))
['cherry', 'banana', 'apple']
CautionChallenge

Searching for a Function

You want to calculate the median of a NumPy array but you do not know the function name. Describe the search query you would use and identify the correct function.

A query like "numpy median array" or browsing the NumPy statistics reference page (numpy.org/doc/stable/reference/routines.statistics.html) will quickly surface numpy.median. You can then confirm its signature with help(numpy.median) or numpy.median?.

TipKey Points
  • Use help(function) or Jupyter’s Shift+Tab / ? shortcuts to read documentation without leaving the notebook.
  • The official NumPy, SciPy, and Matplotlib documentation sites have examples and cross-references that help() does not.
  • Effective searches include the library name, the error type (not the full traceback), and exact function names in quotes.
  • Stack Overflow is reliable for common problems; check the accepted answer and the most-voted answers.
  • When adapting example code, run it as-is first, then change one thing at a time and check the output.