more-itertools

more-itertools extends Python’s standard library itertools module with a large collection of additional iterator-processing functions — chunking, windowing, grouping, peeking, deduplication, flattening, and dozens of other recipes for working with iterables efficiently and lazily. Many of the functions originate from the “itertools recipes” section of the official Python documentation, formalized and battle-tested into a stable public API.

Because it fills gaps left by the standard library without introducing heavyweight dependencies, more-itertools is one of the most widely used utility packages in the Python ecosystem (tens of millions of weekly downloads), commonly pulled in transitively by tools like pip, Sphinx, and other core Python infrastructure.

What You Get

• Chunking and windowing functions (chunked, sliding_window, windowed, ichunked) for processing iterables in fixed-size or overlapping groups
• Peeking and lookahead utilities (peekable, spy, first, last) for inspecting upcoming items without consuming the iterator
• Grouping and partitioning functions (bucket, partition, split_at, split_before, split_after) for splitting a stream by predicate
• Deduplication and set-like operations (unique_everseen, unique_justseen, duplicates_everseen) that preserve original ordering
• Flattening and combining utilities (flatten, collapse, interleave, roundrobin, zip_broadcast) for working with nested or multiple iterables
• A numeric_range generalizing the built-in range() to floats and other numeric types

Common Use Cases

• Data pipeline code that needs to process large or infinite iterables in fixed-size batches without loading everything into memory
• CLI tools and scripts implementing lookahead parsing logic using peekable instead of manually buffering items
• Deduplicating a stream of records while preserving order, without the overhead of sorting or a full set-based approach
• Splitting a sequence into groups by a predicate (e.g. runs of consecutive items) for report generation or log analysis

Under The Hood

Architecture: The library deliberately has just two source modules — more_itertools/recipes.py (1,620 lines, implementing the documented itertools recipes) and more_itertools/more.py (5,557 lines, the larger set of original additions) — both re-exported flatly through more_itertools/__init__.py, so the entire public API is a single flat namespace with no submodule structure to learn. Tech Stack: Pure Python 3.10+ with zero runtime dependencies, built with Flit (flit_core build backend) rather than setuptools, keeping the packaging surface minimal; a requirements/ directory separates dev/test/docs dependency sets. Code Quality: Despite the large function count, the tests/ directory (organized into a handful of test modules covering recipes and more separately) maintains extensive doctest coverage embedded directly in each function’s docstring in addition to standalone unit tests — a pattern that keeps documentation and correctness verification tied together; the project has a long, consistent release history and 40k+ lines of README documenting every function with runnable examples. API Design: Every function follows itertools’ own conventions (lazy iterator returns, composable with itertools functions directly, consistent keyword-argument naming across similar functions like chunked/sliced/windowed), so users already familiar with itertools can adopt new functions with minimal additional learning.