0.4.2

What's new

Added

• Python REPL in the browser demo using pyodide.console.PyodideConsole and jQuery Terminal, with Prism.js syntax highlighting, tab completion, and command history
• 15 curated example query cards (replaced abandoned vaex with python-duckdb, added linting evolution, HTTP clients, and Arrow by platform examples)
• Shareable URL support with a "Copy link" button for reproducible queries
• Quick date range presets (1 month, 3 months, 6 months, 1 year, 2 years)

Changed

• Extract JavaScript and CSS from monolithic index.html into separate app.js and style.css files for maintainability
• Upgrade Pyodide to v0.29.3 and fix API drift (isPyProxy → instanceof pyodide.ffi.PyProxy, to_js import path, echo_command removal)
• Unify "Top 20 packages" into the general query flow instead of a separate code path
• Add prettier pre-commit hook for JS/CSS/HTML formatting

Fixed

• Fix race condition in parallel parquet file loading (fetch in parallel, parse sequentially to avoid global variable collision)
• Fix micropip install by fetching wheel URL from PyPI JSON API
• Fix loading feedback: status displayed on the Query button, active card highlighting for selected examples

Full Changelog: 0.4.1...0.4.2

0.4.1

Fixed

• Fix Pyodide demo to install condastats with deps=False to avoid fsspec version conflict with Pyodide's bundled version

Full changelog: 0.4.0...0.4.1

0.4.0

Added

• New pure-pandas query API: query_overall(), query_grouped(), and top_packages() that operate on any pandas DataFrame without requiring dask or s3fs
• Browser demo now uses the real condastats package installed via micropip from PyPI

Changed

• Refactor _core.py to lazily import dask (only when S3 functions are called), so import condastats works without dask installed (e.g. in Pyodide)
• S3-backed functions now delegate aggregation to the new _query.py module internally

Full changelog: 0.3.0...0.4.0

0.3.0

Breaking Changes

• Drop Python 3.8 and 3.9 support; now requires Python 3.10+
• Add Python 3.12, 3.13, and 3.14 support
• Update to modern dependency versions:
  • numpy>=1.20.0
  • pandas>=2.0.0
  • dask[dataframe]>=2024.5.2
  • pyarrow>=10.0.0

Added

• Interactive browser demo (Pyodide) for trying condastats without installing
• Diátaxis-based documentation with tutorial, how-to guides, explanation, and reference sections
• sphinx-design cards, tabs, and grids on the documentation landing page
• Type hints for all public functions in cli.py
• Comprehensive test suite with 31 tests covering all CLI functions
• Test coverage for multiple packages (pandas, numpy, scipy, requests, dask)
• Dependabot configuration for automated dependency updates
• observed=True parameter to groupby calls (fixes pandas FutureWarning)

Changed

• Switch to conda-sphinx-theme for documentation
• Reorder installer tabs (pixi first), add pixi x run-in-place option
• Migrate from setup.py/versioneer to pyproject.toml/setuptools_scm
• Move tests from condastats/tests/ to top-level tests/ directory
• Update project URLs to conda-incubator organization
• Update ReadTheDocs configuration to use Python 3.11
• Use SPDX license expression (BSD-3-Clause)
• Simplify test fixtures using factory pattern with caching

Removed

• Travis CI configuration (now using GitHub Actions exclusively)
• Local conda recipe (conda-forge feedstock is authoritative)
• requirements.txt (dependencies defined in pyproject.toml)
• setup.py, setup.cfg, versioneer.py, MANIFEST.in

Fixed

• Fix TypeError: descriptor '__call__' on Python 3.11+ by upgrading to dask 2024.5.2+
• Fix ValueError: Not all columns are categoricals (#19) by adding categories=[] to read_parquet
• Fix ArrowStringArray requires PyArrow array of string type (#17, #24) with modern dask/pandas/pyarrow
• Update GitHub Actions to latest versions
• Fix PyPI publish workflow to use supported Python version
• Fix Windows CI environment activation

0.2.1

remove history

0.2.0

use pyarrow instead of fastparquet to solve issue #11