parquet
TLDR
Show file schema
$ parquet-tools schema [file.parquet]
Show metadata$ parquet-tools meta [file.parquet]
Show first rows$ parquet-tools head [file.parquet]
Convert to JSON$ parquet-tools cat --json [file.parquet]
Show row count$ parquet-tools rowcount [file.parquet]
Merge files$ parquet-tools merge [file1.parquet] [file2.parquet] [output.parquet]
SYNOPSIS
parquet-tools command [options] file
DESCRIPTION
Parquet is a columnar storage format for big data. parquet-tools (or parquet-cli) inspects and manipulates Parquet files, showing schema, metadata, and contents.
Parquet provides efficient compression and encoding for analytics workloads.
PARAMETERS
schema
Show schema.meta
Show metadata.head
Show first rows.cat
Output all rows.rowcount
Count rows.merge
Merge files.--json
JSON output.-n num
Number of rows.
PARQUET FEATURES
$ - Columnar storage
- Schema embedded
- Compression (Snappy, GZIP, etc.)
- Predicate pushdown
- Nested data support
- Schema embedded
- Compression (Snappy, GZIP, etc.)
- Predicate pushdown
- Nested data support
PYTHON ALTERNATIVE
$ import pyarrow.parquet as pq
table = pq.read_table('file.parquet')
print(table.schema)
table = pq.read_table('file.parquet')
print(table.schema)
CAVEATS
Java-based tools require JVM. Consider pyarrow for Python workflows. Large files need memory.
HISTORY
Apache Parquet was created as collaboration between Twitter and Cloudera in 2013 for efficient big data storage.
SEE ALSO
orc-tools(1), avro-tools(1), spark(1)
