notebook-cell-tester 1.5.1

Automatic grading framework for student code in Google Colab and Jupyter notebooks

notebook-cell-tester

Automatic grading framework for student code in Google Colab and Jupyter notebooks.

Instructors drop a single test cell below the student's code cell. The framework reads the previous cell's source from IPython history, executes it, runs the configured tests, and renders a color-coded HTML results table — no extra setup required from students.

---

Table of contents

• Installation
• Quick start
• How it works
• Test types
  • return — function return value
  • output — exact stdout match
  • exception — expected exception
  • regex / not_regex — source code pattern
  • variable — value with validator
  • type_check — isinstance check
  • partial_output — fuzzy match
  • regex_output — stdout regex
  • contains_output — stdout substring
  • multiline_output — order-independent lines
• Sections
• TestCase field reference
• Results display
• Common patterns
• License

---

Installation

pip install notebook-cell-tester

Colab one-liner — paste this at the top of the test cell if the package is not pre-installed in the environment:

!pip install -q notebook-cell-tester

---

Quick start

Place this cell immediately after the student's code cell in the notebook:

from notebook_cell_tester import ColabTestFramework, TestCase

tester = ColabTestFramework()

tests = [
    TestCase(
        name="add_numbers returns correct sum",
        test_type="return",
        function_name="add_numbers",
        inputs=[3, 4],
        expected=7,
        description="add_numbers(3, 4) should return 7",
    ),
    TestCase(
        name="Prints greeting",
        test_type="output",
        stdin_input="Alice",
        expected="Hello, Alice!",
    ),
]

tester.run_tests(tests)
tester.display_results()

Running this cell produces a color-coded HTML table showing each test's pass/fail status, expected vs. actual values, and (when something crashes) a collapsible traceback that students can expand.

---

How it works

1. ColabTestFramework() is instantiated in the test cell — the cell below the student's code.
2. run_tests(tests) calls load_last_cell(), which reads In[-2] from the IPython namespace to get the second-to-last cell (the student's code, not the test cell).
3. Each TestCase is dispatched to the appropriate internal method based on test_type.
4. display_results() renders an HTML table with pass/fail badges, result details, and collapsible error tracebacks.

Tests that only inspect the IPython namespace (variable, type_check on a variable) run even when the student hasn't executed their cell yet.

---

Test types

Summary

test_type	Checks	Key fields
return	Function return value	function_name, inputs, expected, tolerance
output	Exact printed output	stdin_input, expected, optional function_name
exception	Function raises expected exception	function_name, inputs, expected
regex	Source code matches pattern	pattern
not_regex	Source code does NOT match pattern	pattern
variable	Variable satisfies validator	variable_name, validator
type_check	Value is correct type	function_name or variable_name, expected
partial_output	Output similarity ≥ threshold (Levenshtein)	expected, similarity_threshold
regex_output	Printed output matches regex	pattern
contains_output	Printed output contains substring	expected
multiline_output	Every expected line appears in output	expected

---

return — function return value

Calls function_name(*inputs) and checks that the return value equals expected.

TestCase(
    name="Square of 5",
    test_type="return",
    function_name="square",
    inputs=[5],
    expected=25,
)

Use tolerance for floating-point comparisons (abs(result - expected) <= tolerance):

TestCase(
    name="Average of [1, 2, 3]",
    test_type="return",
    function_name="average",
    inputs=[[1, 2, 3]],
    expected=2.0,
    tolerance=1e-9,
)

---

output — exact stdout match

Captures everything printed to stdout and compares it to expected (after stripping leading/trailing whitespace).

# Test the whole cell — no function_name needed
TestCase(
    name="Greeting output",
    test_type="output",
    stdin_input="Alice",
    expected="Hello, Alice!",
)

# Test a specific function's printed output
TestCase(
    name="print_square(4)",
    test_type="output",
    function_name="print_square",
    inputs=[4],
    expected="16",
)

Supply stdin_input to simulate one or more input() calls. Separate multiple lines with \n.

---

exception — expected exception

Passes when function_name(*inputs) raises exactly the expected exception type. Fails if it returns normally or raises a different exception.

TestCase(
    name="Division by zero raises ValueError",
    test_type="exception",
    function_name="safe_divide",
    inputs=[10, 0],
    expected=ValueError,
)

---

regex / not_regex — source code pattern

Searches the student's source code for a Python regex. Useful for requiring or forbidding specific language constructs without running the code.

# Require a for loop
TestCase(
    name="Uses a for loop",
    test_type="regex",
    pattern=r"\bfor\b",
    description="Your solution must use a for loop.",
)

# Forbid the built-in sum()
TestCase(
    name="Does not call sum()",
    test_type="not_regex",
    pattern=r"\bsum\s*\(",
    error_message="Implement the summation manually — do not use sum().",
)

Both types use re.MULTILINE | re.DOTALL flags.

---

variable — value with validator

Reads a variable from the IPython namespace and passes it to a validator callable. The test passes when validator(value) returns True.

TestCase(
    name="score is between 0 and 100",
    test_type="variable",
    variable_name="score",
    validator=lambda v: 0 <= v <= 100,
    error_message="score = {value}, but it must be between 0 and 100.",
)

Use {value} as a placeholder in error_message — it is replaced with the actual variable value when the test fails.

---

type_check — isinstance check

Passes when isinstance(value, expected) is True. Works with a function's return value or a named variable.

# Check a function's return type
TestCase(
    name="get_scores() returns a list",
    test_type="type_check",
    function_name="get_scores",
    inputs=[],
    expected=list,
)

# Check a variable's type
TestCase(
    name="result is a float",
    test_type="type_check",
    variable_name="result",
    expected=float,
)

# Accept multiple types using a tuple
TestCase(
    name="index is int or float",
    test_type="type_check",
    variable_name="index",
    expected=(int, float),
)

---

partial_output — fuzzy match

Passes when the Levenshtein similarity between the actual and expected output is ≥ similarity_threshold. Useful when minor formatting differences (extra spaces, punctuation) should still receive credit.

TestCase(
    name="Greeting (fuzzy match)",
    test_type="partial_output",
    stdin_input="Bob",
    expected="Hello, Bob! Welcome.",
    similarity_threshold=0.85,
    description="At least 85% similar to the expected greeting.",
)

similarity_threshold must be a float in (0.0, 1.0]. A value of 1.0 requires an exact match; 0.8 allows up to ~20% edit distance.

---

regex_output — stdout regex

Passes when re.search(pattern, output) finds a match in the printed output. Good for checking output format without requiring an exact string.

TestCase(
    name="Output contains a decimal number",
    test_type="regex_output",
    pattern=r"\d+\.\d+",
    error_message="Expected a decimal number in the output.",
)

# With a specific function
TestCase(
    name="format_price() output includes currency symbol",
    test_type="regex_output",
    function_name="format_price",
    inputs=[9.99],
    pattern=r"\$\d+\.\d{2}",
)

---

contains_output — stdout substring

Passes when expected is a substring of the printed output (case-sensitive).

TestCase(
    name="Output mentions 'Error'",
    test_type="contains_output",
    expected="Error",
    description="The program must print an error message.",
)

---

multiline_output — order-independent lines

Splits expected on newlines and verifies that every non-empty line appears somewhere in the output. Order does not matter — useful when students may print items in any sequence.

TestCase(
    name="Prints name and age",
    test_type="multiline_output",
    expected="Name: Alice\nAge: 30",
    description="Both lines must appear in the output (any order).",
)

---

Sections

Use TestSection to group tests into named blocks, each rendered as a separate table with its own header. Import it alongside ColabTestFramework and TestCase, then call run_sections() instead of run_tests().

from notebook_cell_tester import ColabTestFramework, TestCase, TestSection

tester = ColabTestFramework()

tester.run_sections([
    TestSection("Part 1: Code structure", [
        TestCase(
            name="Uses a for loop",
            test_type="regex",
            pattern=r"\bfor\b",
            description="Your solution must iterate with a for loop.",
        ),
        TestCase(
            name="Does not use sum()",
            test_type="not_regex",
            pattern=r"\bsum\s*\(",
            error_message="Implement the summation manually — do not call sum().",
        ),
    ]),
    TestSection("Part 2: Correctness", [
        TestCase(
            name="total([1, 2, 3]) == 6",
            test_type="return",
            function_name="total",
            inputs=[[1, 2, 3]],
            expected=6,
        ),
        TestCase(
            name="total([]) == 0",
            test_type="return",
            function_name="total",
            inputs=[[]],
            expected=0,
        ),
    ]),
])

tester.display_results()

display_results() automatically detects that sections were used and renders a named header bar above each group showing the section title and its own n/total passed count, followed by an overall summary banner at the top.

run_tests() with a flat list continues to work exactly as before — sections are opt-in.

---

TestCase field reference

Field	Type	Required for	Notes
name	str	all	Display name shown in the results table
test_type	str	all	One of the test type strings above
function_name	str	function tests, type_check (one of)	Looked up from get_ipython().user_ns
variable_name	str	variable, type_check (one of)	Looked up from get_ipython().user_ns
inputs	list	function tests	Positional arguments; defaults to []
stdin_input	str	output tests	Simulates input() calls; lines separated by \n
expected	Any	most tests	Return value, output string, exception type, or target type
similarity_threshold	float	partial_output	Must be in (0.0, 1.0]
tolerance	float	return (optional)	Passes when abs(got - expected) <= tolerance; must be >= 0
validator	Callable	variable	lambda value: bool
pattern	str	regex, not_regex, regex_output	Python regex string
description	str	optional on all	Shown as an italic subtitle under the test name
error_message	str	optional on all	Custom failure message; use {value} placeholder in variable tests
success_message	str	optional on all	Custom message shown when the test passes; replaces the default pass message

---

Results display

display_results() renders an HTML table inside the notebook:

• Summary banner — gradient bar showing passed / total (%) with a congratulatory message when all tests pass.
• Section headers (when using run_sections()) — a purple gradient bar above each group showing the section name and its own n/total passed count.
• Status column — green ✓ PASS or red ✗ FAIL badge per test row.
• Test column — test name, with description shown as a small italic subtitle when set.
• Details column — human-readable message (expected vs. actual values, similarity percentages, missing lines, etc.).
• Collapsible traceback — when an unhandled exception occurs during execution, a ▶ Show technical details disclosure element hides the raw traceback by default, keeping the table readable for beginners.

---

Common patterns

Test an entire cell (no function)

Omit function_name. The framework re-executes the student's full cell source in an isolated namespace with stdin redirected.

TestCase(
    name="Reads two numbers and prints their sum",
    test_type="output",
    stdin_input="3\n7",
    expected="10",
)

Combine structural and functional tests

tests = [
    # 1. Check the algorithm used
    TestCase(
        name="Uses recursion",
        test_type="regex",
        pattern=r"def\s+factorial.*?\n(?:.*\n)*?.*\bfactorial\s*\(",
        description="factorial() must call itself recursively.",
    ),
    # 2. Check correctness
    TestCase(
        name="factorial(5) == 120",
        test_type="return",
        function_name="factorial",
        inputs=[5],
        expected=120,
    ),
    # 3. Check error handling
    TestCase(
        name="Raises ValueError for negative input",
        test_type="exception",
        function_name="factorial",
        inputs=[-1],
        expected=ValueError,
    ),
]

Check a variable's value and type together

tests = [
    TestCase(
        name="result is a float",
        test_type="type_check",
        variable_name="result",
        expected=float,
    ),
    TestCase(
        name="result is positive",
        test_type="variable",
        variable_name="result",
        validator=lambda v: v > 0,
        error_message="result = {value}, but it must be greater than 0.",
    ),
]

Custom pass and fail messages

Use success_message to replace the default technical pass message with something more encouraging, and error_message to give students a clear hint on failure:

TestCase(
    name="factorial(5) == 120",
    test_type="return",
    function_name="factorial",
    inputs=[5],
    expected=120,
    success_message="Correct! Your factorial function works perfectly.",
    error_message="Check your base case and recursive step.",
)

---

Flexible output with partial_output

When students may phrase output slightly differently, use partial_output to award credit for near-correct answers:

TestCase(
    name="Farewell message",
    test_type="partial_output",
    stdin_input="Maria",
    expected="Goodbye, Maria! See you soon.",
    similarity_threshold=0.80,
    description="Your farewell message must be at least 80% similar to the expected output.",
)

---

License

MIT — see LICENSE.