pytae

Fast, ergonomic Pandas tools and zero-code CLI for tabular data manipulation, feature engineering, DuckDB SQL, and visualization.

pytae provides two complementary ways to work with tabular data:

Python Library & DataFrame Accessor (df.pt): Ergonomic data-science verbs (qry, select, mutate, agg, long, wide, pivot, sql, plot) that compose smoothly with vanilla Pandas.
Unix Pipeline CLI (pytae): Inspect, filter, derive columns, aggregate, join, diff, and convert tabular files (.parquet, .csv, .tsv, .jsonl, .dat, .sas7bdat, .gz) directly from the terminal without writing Python code.

Installation
Quick Start: CLI
Quick Start: Python Library
Key Features
Key Syntax & Conventions Cheat Sheet
Documentation & Interactive Tutorials
License

Installation

pip install pytae              # Core library and CLI
pip install "pytae[plot]"       # Includes matplotlib & scipy for plotting
pip install "pytae[sql]"        # Includes DuckDB for SQL queries
pip install "pytae[plot,sql]"   # All optional dependencies

Quick Start: CLI

pytae commands execute in exact terminal order like a Unix pipeline:

# 1. Fast inspection: view shape, schema, and first 5 rows without reading full files
pytae penguins.parquet -shape -head 5

# 2. Filter, engineer features, sort, and export to CSV
pytae penguins.parquet \
  -qry "species = 'Adelie', body_mass_g > 3500" \
  -mutate "mass_kg = body_mass_g / 1000, ratio = bill_length_mm / bill_depth_mm" \
  -sort_by "mass_kg desc" \
  -select "species,island,mass_kg,ratio" \
  -o heavy_adelie.csv

# 3. Grouped window calculation: calculate group mean and deviation per row
pytae tips.parquet \
  -by day \
  -mutate "avg_tip = mean(tip), diff = tip - avg_tip, sz = n" \
  -round 2 \
  -head 5

# 4. Grouped aggregation and pivoting
pytae penguins.parquet \
  -by 'island,species' \
  -agg 'avg_mass=body_mass_g:mean, count=n' \
  -wide 'c=species, v=avg_mass' \
  -round 1

# 5. Query any dataset using DuckDB SQL
pytae sales.parquet -sql "select customer_id, sum(total) as revenue from data group by 1 order by 2 desc" -head 10

# 6. Compare schemas and values between two files
pytae current.parquet -diff previous.parquet

# 7. In-line terminal visualizations & ASCII charts
pytae penguins.parquet -freq species
pytae penguins.parquet -hist body_mass_g:10

# 8. Render and export figures directly from the CLI
pytae penguins.parquet -by species -agg "body_mass_g=mean" -plot "kind=bar, x=species, y=body_mass_g" -o mass_chart.png

# 9. Pipe streaming via standard input (STDIN)
cat penguins.csv | pytae - -select "species,island" -value_counts

👉 Full CLI Guide: docs/cli.md | Individual Feature Guides: docs/cli/

Quick Start: Python Library

Use verbs as standalone functions (pt.verb(df, ...)) or chain them naturally on any DataFrame via the .pt accessor (df.pt.verb(...)):

import pytae as pt
import pandas as pd

# Load sample dataset
penguins = pt.sample("penguins")

# Clean, expressive pipeline
summary = (
    penguins
    # 1. Row filtering with comparison and intervals
    .pt.qry(species=["Adelie", "Gentoo"], body_mass_g=">= 3500")
    # 2. Grouped window calculation (N -> N)
    .pt.mutate(
        mass_kg="body_mass_g / 1000",
        avg_mass="mean(body_mass_g)",
        diff="body_mass_g - avg_mass",
        by="species",
    )
    # 3. Select columns with negative drop support
    .pt.select("species", "island", "mass_kg", "diff")
    # 4. Grouped aggregation (N -> K)
    .pt.agg(by="species", mean_mass="mass_kg.mean()", count="n")
)

print(summary)

👉 Full Library Guide: docs/library.md | Interactive Tutorials: docs/library/

Key Features

1. Row Filtering (`qry`)

Filter rows without boilerplate. Accepts Python keywords, string expressions, and mathematical intervals:

# Keywords, lists, and closed intervals [3000, 4000]
df.pt.qry(species=["Adelie", "Gentoo"], body_mass_g="[3500, 4500]")

# Columns containing spaces use square brackets
df.pt.qry("[bill length mm] > 40")

# Missing value checks
df.pt.qry(sex=("notna",))

2. Column Creation & Grouped Window Calculations (`mutate`)

Derive columns with Python/Pandas expressions, dplyr-style conditionals (if_else, case_when, coalesce), and grouped window summaries:

# Formulas, vectorized conditionals, and caller-scope variables (@threshold)
threshold = 4000
df.pt.mutate(
    mass_kg="body_mass_g / 1000",
    size_class="if_else(body_mass_g > @threshold, 'Large', 'Normal')",
)

# Grouped window calculations: aggregates evaluate per group and broadcast back (N -> N)
df.pt.mutate(
    avg_mass="mean(body_mass_g)",
    diff="body_mass_g - avg_mass",
    group_size="n",
    by="species",
)

3. Column Selection & Exclusion (`select`)

Reorder, slice, and filter columns using names, patterns, types, and negative drops:

# Negative selection (drop specific columns)
df.pt.select("-id", "-temp")

# By pattern, dtype, and position
df.pt.select("species", contains="bill", dtypes="numeric")

# Reorder with everything helper
df.pt.select("island", "species", pt.everything)

4. Grouped Aggregation (`agg`)

Concise group summaries with explicit by= grouping, bracket support for spaced names, and automatic row count token n:

# Explicit grouping and named aggregations
df.pt.agg(by="species", avg_mass="body_mass_g.mean()", n="n")

# String mapping specification (perfect for spaced columns)
df.pt.agg("island", "[avg mass] = [body mass g]:mean, n = n")

5. Pure Reshaping (`long` & `wide`)

Clean, deterministic unpivoting (melting) and spreading without MultiIndex complexity:

# 1. Unpivot non-numeric columns into key-value pairs
long_df = df.pt.long(id_vars=["species", "island"], cols=["bill_length_mm", "bill_depth_mm"])

# 2. Pure reshape back to wide presentation matrix
wide_df = long_df.pt.wide(c="variable", v="value")

6. 2D Pivot Tables (`pivot`)

Excel-style 2D pivot tables with automatic index reset and flat 1D columns:

# 2D grid summarizing average body mass across island and species
pivot_df = df.pt.pivot(r="island", c="species", v="body_mass_g", a="mean")

# Frequency matrix (count rows across dimensions; v is optional for a="n")
freq_matrix = df.pt.pivot(r="island", c="species", a="n")

7. Embedded DuckDB SQL Engine (`sql`)

Run analytical SQL directly on any Pandas DataFrame with zero copy:

result = df.pt.sql("""
    select species, count(*) as total, round(avg(body_mass_g), 1) as avg_mass
    from data
    where body_mass_g is not null
    group by 1
    order by avg_mass desc
""")

8. Visualization (`plot`)

Method-chainable charting powered by Matplotlib (pip install "pytae[plot]"):

# Direct accessor chaining:
penguins.pt.plot(
    kind="scatter",
    x="bill_length_mm",
    y="bill_depth_mm",
    c="species",
    cmap="viridis",
    title="Penguin Bill Dimensions",
).pt.finalize()

# One-shot direct file export (no explicit finalize required):
penguins.pt.agg("species", body_mass_g="mean").pt.plot(kind="bar", x="species", y="body_mass_g").save("mass.png")

9. Zero-Cost Metadata Inspection & Diffing

Inspect large Parquet, CSV, and SAS datasets instantly without loading millions of rows into memory:

# Instant shape, columns, and types from Parquet file headers
pytae large_dataset.parquet -shape -cols -meta

# Compare schemas, null counts, and value differences across datasets
pytae current.parquet -diff previous.parquet

Key Syntax & Conventions Cheat Sheet

Task	Python Library (`df.pt`)	CLI Flag	Example
Row Filtering	`.pt.qry(...)`	`-qry "..."`	`df.pt.qry(species="Adelie", mass="> 3500")`
Spaced Columns (Expr)	`[column name]`	`[column name]`	`df.pt.mutate(ratio="[bill length mm] / [bill depth mm]")`
Grouped Window Calc	`.pt.mutate(..., by=...)`	`-by ... -mutate "..."`	`pytae data.parquet -by dept -mutate "avg=mean(salary)"`
Negative Column Select	`.pt.select("-col1", "-col2")`	`-select "-col1,-col2"`	`df.pt.select("-temp", "-raw_id")`
2D Pivot Table	`.pt.pivot(...)`	`-pivot "..."`	`df.pt.pivot(r="Region", c="Year", v="Sales", a="sum")`
Data Reshaping	`.pt.long()`, `.pt.wide()`	`-long`, `-wide`	`df.pt.wide(c="metric", v="val")`
In-Line Terminal Charts	N/A	`-freq`, `-hist`	`pytae data.parquet -freq species` / `-hist mass:10`
Figure Plotting & Export	`.pt.plot(..., save=...)`	`-plot "..." -o ...`	`pytae data.parquet -plot "kind=bar, x=sp, y=wt" -o out.png`
DuckDB SQL Query	`.pt.sql("select ...")`	`-sql "select ..."`	`pytae data.parquet -sql "select * from data limit 5"`
Pipe / STDIN Stream	N/A	`pytae - [-fmt ...]`	`cat data.csv \| pytae - -head 5`
Copy to Clipboard	`df.to_clip()`	`-o clip`	`df.head().to_clip()` vs `pytae data.parquet -head -o clip`
File Export & Routing	`df.to_parquet(...)`	`-o <target.ext>`	`pytae data.parquet -o clean.csv.gz -od ./exports`
Dataset Diffing	N/A	`-diff <other_file>`	`pytae data.parquet -diff old_data.parquet`

Documentation & Interactive Tutorials

CLI Documentation (`docs/cli/`)

CLI Reference Hub — Master flag guide, execution pipeline, and syntax rules
Aggregation Guide — -by, -agg, and group frequency counts
Mutation Guide — Feature engineering, math, and grouped window formulas
Filtering Guide — Numerical comparisons, string matching, and intervals
Selection Guide — Column slicing, dtypes, and negative exclusion
Pure Reshaping Guide — Long-to-wide and wide-to-long pure reshapes (-long, -wide)
2D Pivot Tables Guide — Excel-style 2D pivot tables with automatic index reset (-pivot)
Visualizations & Plotting Guide — In-line terminal charts (-freq, -hist) and figure export (-plot)
Export & Output Guide — -o, -od, compression, and clipboard routing
Dataset Diff Guide — Comparing schemas and records
Multi-File Operations — -file, -merge, and -concat
SQL Guide — Querying with embedded DuckDB
Metadata & Inspection Guide — -shape, -cols, -meta, and zero-cost header inspection
Cleaning & Replacement Guide — -clean_columns and -replace_values

Interactive Jupyter Notebooks (`docs/library/`)

Each core module has a standalone, fully runnable tutorial with live outputs:

Filtering: docs/library/qry.ipynb
Selection: docs/library/select.ipynb
Feature Engineering & Window Mutations: docs/library/mutate.ipynb
Aggregation: docs/library/agg.ipynb
Pure Reshaping: docs/library/reshape.ipynb
2D Pivot Tables: docs/library/pivot.ipynb
DuckDB SQL: docs/library/sql.ipynb
Plotting & Dashboards: docs/library/plotting.ipynb
Utilities & Cleaning: docs/library/other_utilities.ipynb

License

MIT License. See LICENSE for details.

Metadata

Release files for pytae 3.7.2

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for pytae 3.7.2
File	Size	Uploaded	Provenance
pytae-3.7.2.tar.gz	3.1 MB	Oct 4, 2026	Without provenance	Details

Built distribution (wheel)

Table of built distributions (wheels) for pytae 3.7.2
File	Interpreter	ABI	Platform	Reset
pytae-3.7.2-py3-none-any.whl 1.9 MB Oct 4, 2026 Without provenance	Python 3	none	any	Details

Total release size: 5.0 MB

Release files / pytae-3.7.2.tar.gz

Download URL	pytae-3.7.2.tar.gz
Size	3.1 MB
Tags	Source
SHA-256 checksum How to use checksums	`0b4549a25bb51f1ed474c2d9ff20ef0720384b5650adc6e3aa23c2b2999db8f5`
BLAKE2b-256 checksum How to use checksums	`68d68748ba58c6ae6828f19158de136cb65c31c9e19f58b85ca7f49fcaaeccb7`
Upload date	Oct 4, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Release files / pytae-3.7.2-py3-none-any.whl

Download URL	pytae-3.7.2-py3-none-any.whl
Size	1.9 MB
Tags	Python 3
SHA-256 checksum How to use checksums	`faed5addbcda0b60d0cd1cc35dcd30d88260e8b27a97fa274940f6d00736cce1`
BLAKE2b-256 checksum How to use checksums	`000805021091f15f262f97d2de725e750be8ac36eece312d0e90372fe895653b`
Upload date	Oct 4, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Security for pytae 3.7.2

Report project as malware

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.