pqfilt 0.1.4

Generic Parquet filtering tool (CLI + API)

pqfilt

Generic Parquet filtering tool (CLI and Python API).

ReadtheDocs Documentation.

Originally developed while dealing with large Parquet files in SPHEREx mission (GitHub).

pqfilt wraps pyarrow.dataset to let you filter Parquet files before they are fully read into memory, using row-group-level filtering.

Installation

pip install pqfilt
# or
uv add pqfilt

Python API

import pqfilt

# Simple filter
df = pqfilt.read("data.parquet", filters="vmag < 20")

# AND + OR with expression syntax
df = pqfilt.read("data.parquet", filters="(a < 30 & b > 50) | c == 1")

# Membership filter (explicit quotes preserve string types, e.g., to prevent Parquet type errors)
df = pqfilt.read("data.parquet", filters="desig in '3200', '356', '134'")

# Tuple syntax (flat AND)
df = pqfilt.read("data.parquet", filters=[("a", "<", 30), ("b", ">", 50)])

# DNF syntax (OR of ANDs)
df = pqfilt.read("data.parquet", filters=[
    [("a", "<", 30)],
    [("b", ">", 50)],
])

# Column selection + output
df = pqfilt.read("data/*.parquet", columns=["a", "b"], output="out.parquet")

CLI

# Basic filter
pqfilt data/*.parquet -f "vmag < 20" -o filtered.parquet

# AND + OR expression
pqfilt data/*.parquet -f "(a < 30 & b > 50) | c == 1" -o filtered.parquet

# Multiple -f flags (AND-ed together)
pqfilt data/*.parquet -f "vmag < 20" -f "dec > 30" -o filtered.parquet

# Column selection
pqfilt data/*.parquet -f "vmag < 20" --columns vmag,ra,dec -o filtered.parquet

# Membership filter
pqfilt data/*.parquet -f "desig in 1,2,3" -o filtered.parquet

Column names with special characters

Columns containing operator characters can be backtick-quoted:

pqfilt.read("data.parquet", filters="`alpha*360` > 100")

License

MIT