filequery

Query CSV, JSON and Parquet files using SQL.

• runs queries using a DuckDB in-memory database for efficient querying
• any SQL that works with DuckDB will work here
• use the CLI to easily query files in your terminal or automate queries/transformations as part of a script
• use the TUI for a more interactive experience

Demo

CLI

TUI

Installation

pipx install filequery

or

pip install filequery

CLI usage

Run filequery --help to see what options are available.

usage: filequery [-h] [-f FILENAME] [-d FILESDIR] [-q QUERY] [-Q QUERY_FILE] [-o OUT_FILE [OUT_FILE ...]] [-F OUT_FILE_FORMAT] [-D DELIMITER] [-c CONFIG] [-e] [-v]
options:
-h, --help show this help message and exit
-f FILENAME, --filename FILENAME
path to a CSV, Parquet or JSON file
-d FILESDIR, --filesdir FILESDIR
path to a directory which can contain a combination of CSV, Parquet and JSON files
-q QUERY, --query QUERY
SQL query to execute against file
-Q QUERY_FILE, --query_file QUERY_FILE
path to file with query to execute
-o OUT_FILE [OUT_FILE ...], --out_file OUT_FILE [OUT_FILE ...]
file to write results to instead of printing to standard output
-F OUT_FILE_FORMAT, --out_file_format OUT_FILE_FORMAT
either csv or parquet, defaults to csv
-D DELIMITER, --delimiter DELIMITER
delimiter to use when printing result or writing to CSV file
-c CONFIG, --config CONFIG
path to JSON config file
-e, --editor run SQL editor UI for exploring data
-v, --version show program's version number and exit

For basic usage, provide a path to a CSV or Parquet file and a query to execute against it. The table name will be the file name without the extension. If the file name does not conform to DuckDB's rules for unquoted identifiers, the table name will need to be wrapped in double quotes. For example, a file named my data.csv would be queried as select * from "my data".

filequery --filename example/test.csv --query 'select * from test'

TUI usage

To use the TUI for querying your files, use the -e flag and provide a path to a file or directory.

filequery -e -f path/to/file.csv

or

filequery -e -d path/to/file_directory

You can also omit a path to a file or directory and open a blank editor. This can be helpful if you want to directly use DuckDB functions such as read_csv_auto() for querying your files.

filequery -e

Examples

filequery --filename example/json_test.json --query 'select nested.nest_id, nested.nest_val from json_test'# query json

filequery --filesdir example/data --query 'select * from test inner join test1 on test.col1 = test1.col1'# query multiple files in a directory

filequery --filesdir example/data --query_file example/queries/join.sql # point to a file containing SQL

filequery --filesdir example/data --query_file example/queries/json_csv_join.sql # SQL file joining data from JSON and CSV files

filequery --filesdir example/test.csv --query 'select * from test; select sum(col3) from test;'# output multiple query results to multiple files

filequery --filename example/ndjson_test.ndjson --query 'select id, value, nested.subid, nested.subval from ndjson_test'# query nested JSON in an ndjson file

You can also provide a config file instead of specifying the arguments when running the command.

filequery --config <path to config file>

The config file should be a json file. See example config file contents below.

{
"filename": "../example/test.csv",
"query": "select col1, col2 from test"
}

{
"filesdir": "../example/data",
"query_file": "../example/queries/join.sql",
"out_file": "result.parquet",
"out_file_format": "parquet"
}

See the example directory in the repo for more examples.

Module usage

You can also use filequery in your own programs. See the example below.

fromfilequery.filedbimportFileDbquery='select * from test'# read test.csv into a table called "test"fdb=FileDb('example/test.csv')
# return QueryResult objectres=fdb.exec_query(query)
# formats result as csvprint(str(res))
# saves query result to result.csvres.save_to_file('result.csv')
# saves query result as parquet filefdb.export_query(query, 'result.parquet', FileType.PARQUET)

Development

Packages required for distribution should go in requirements.txt.

To build the wheel:

pip install -r requirements-dev.txt
make

Testing

To test the CLI, create a separate virtual environment perform an editable install.

python -m venv test-env
. test-env/bin/activate
pip install -e .

To run unit tests, stay in the root of the project. The unit tests add src to the path so filequery can be imported properly.

python tests/test_filequery.py