statistics: covariance, correlation, and linear_regression do not accept iterator inputs

[htjworld]

Bug description

statistics.covariance(), statistics.correlation(), and
statistics.linear_regression() raise TypeError when given iterator or
generator inputs, unlike the rest of the statistics module.

Reproduction

import statistics

# These functions accept iterators without issue:
statistics.mean(iter([1, 2, 3]))         # 2
statistics.variance(iter([1, 2, 3, 4]))  # 1.666...
statistics.stdev(iter([1, 2, 3, 4]))     # 1.290...

# These three raise TypeError:
statistics.covariance(iter([1, 2, 3, 4, 5]), iter([2, 4, 6, 8, 10]))
# TypeError: object of type 'list_iterator' has no len()

statistics.correlation(iter([1, 2, 3, 4, 5]), iter([2, 4, 6, 8, 10]))
# TypeError: object of type 'list_iterator' has no len()

statistics.linear_regression(iter([1, 2, 3, 4, 5]), iter([2, 4, 6, 8, 10]))
# TypeError: object of type 'list_iterator' has no len()

Expected behavior

All three functions should accept any iterable (lists, tuples, iterators,
generators), consistent with mean(), variance(), and stdev().

The private _ss() helper — used internally by variance() and stdev() —
documents this design intent explicitly:

"Calculations are done in a single pass, allowing the input to be an iterator."

Root cause

All three functions call len(x) on the raw input at the top of the
function body, which fails for iterators that do not support len().

CPython version

Reproducible on main.

Linked PRs

• gh-149244: Support iterator inputs in covariance, correlation, and linear_regression #149245
• gh-149244 Document statistics functions that require sequence inputs. #149264

[rhettinger]

The LLM falsely assumed that because the _ss helper function handled iterator inputs that all statistics functions should as well. However, that was not the intended behavior for covariance, correlation, and linear_regression.

I will make a doc edit to document the intended behavior.

I don't think these functions benefit from adding list(it) which would do an unnecessary copy in the common case.

[htjworld]

Thanks for the explanation and the doc fix! Makes sense that the bivariate functions intentionally require sequences.