A blistering-fast Avro CLI tool — a modern replacement for Java's avro-tools and Python's fastavro.
Supports local files, glob patterns, and S3 URIs.
brew tap arunma/tap
brew install avzcargo install avzDownload the .deb from Releases:
sudo dpkg -i avz-x86_64-unknown-linux-gnu.debgit clone https://github.com/arunma/avz.git
cd avz
cargo build --release
# binary at target/release/avz# peek at the first 5 records
avz head -n 5 data.avro
# pretty-print with syntax highlighting
avz cat --pretty data.avro
# search for a record by regex
avz grep "user_id.*12345" data.avro
# search for a literal string (no regex)
avz grep -F "300376641*2967" data.avro
# count records across files using a glob
avz count "logs/*.avro"
# works with S3 too
avz cat "s3://my-bucket/events/dt=2026-03-16/*.avro" --prettyNote: Quote glob patterns and S3 URIs to prevent your shell from expanding them.
Usage: avz <COMMAND>
Commands:
cat Print all records as JSON
head Print the first N records (default 10)
schema Print the Avro schema as JSON
count Count records in Avro files
meta Print file metadata (codec, sync marker, user metadata)
fromjson Convert JSON records to an Avro file
concat Concatenate Avro files into one
recodec Re-encode an Avro file with a different codec
fingerprint Print schema fingerprint (CRC-64-AVRO, MD5, SHA-256)
validate Validate an Avro file or check schema compatibility
grep Search records for a pattern, printing matching records as JSON
random Generate random records from a schema
All examples below use this sample dataset — 8 employee records with fields for id, name, department (enum), salary, email (nullable), tags (array), and active (boolean).
Print all records, one JSON object per line:
$ avz cat employees.avro{"id":1,"name":"Alice Chen","department":"ENGINEERING","salary":125000.0,"email":"alice@example.com","tags":["rust","backend","senior"],"active":true}
{"id":2,"name":"Bob Smith","department":"SALES","salary":95000.0,"email":"bob@example.com","tags":["enterprise","closer"],"active":true}
{"id":3,"name":"Carol Davis","department":"ENGINEERING","salary":140000.0,"email":"carol@example.com","tags":["rust","systems","principal"],"active":true}
...With --pretty for colorized, indented output:
$ avz cat --pretty employees.avro{
"id": 1,
"name": "Alice Chen",
"department": "ENGINEERING",
"salary": 125000.0,
"email": "alice@example.com",
"tags": [
"rust",
"backend",
"senior"
],
"active": true
}Limit output with -n:
$ avz cat -n 2 employees.avro$ avz head -n 3 employees.avro{"id":1,"name":"Alice Chen","department":"ENGINEERING","salary":125000.0,"email":"alice@example.com","tags":["rust","backend","senior"],"active":true}
{"id":2,"name":"Bob Smith","department":"SALES","salary":95000.0,"email":"bob@example.com","tags":["enterprise","closer"],"active":true}
{"id":3,"name":"Carol Davis","department":"ENGINEERING","salary":140000.0,"email":"carol@example.com","tags":["rust","systems","principal"],"active":true}With colorized output:
$ avz head -n 2 --pretty employees.avro{
"id": 1,
"name": "Alice Chen",
"department": "ENGINEERING",
"salary": 125000.0,
"email": "alice@example.com",
"tags": [
"rust",
"backend",
"senior"
],
"active": true
}
{
"id": 2,
"name": "Bob Smith",
"department": "SALES",
"salary": 95000.0,
"email": "bob@example.com",
"tags": [
"enterprise",
"closer"
],
"active": true
}Default is 10 records when -n is omitted.
Outputs colorized JSON with automatic pager for large schemas:
$ avz schema employees.avro{
"name": "com.example.hr.Employee",
"type": "record",
"fields": [
{
"name": "id",
"type": "int"
},
{
"name": "name",
"type": "string"
},
{
"name": "department",
"type": {
"name": "com.example.hr.Department",
"type": "enum",
"symbols": [
"ENGINEERING",
"SALES",
"MARKETING",
"HR",
"FINANCE"
]
}
},
{
"name": "salary",
"type": "double"
},
{
"name": "email",
"type": [
"null",
"string"
]
},
{
"name": "tags",
"type": {
"type": "array",
"items": "string"
}
},
{
"name": "active",
"type": "boolean"
}
]
}Large schemas automatically pipe through less -R in interactive terminals.
Single file:
$ avz count employees.avro
8Multiple files show per-file counts and a total:
$ avz count employees.avro employees2.avro
employees.avro: 8
employees2.avro: 8
total: 16Works with globs:
$ avz count "data/*.avro"Shows the raw schema, codec, sync marker, and any user-defined metadata:
$ avz meta employees.avroavro.schema { ... }
avro.codec null
sync 0be4e3b6562329dbba6c5f06aa43ee96
Print all fingerprints:
$ avz fingerprint employees.avro
CRC-64-AVRO 146d06fde15d172f
MD5 874856ac6f65f6eeced12661790a5ec2
SHA-256 3c9dd71e34662cb613aac0d4bdb7afa7309f2712ff97c1991a29028fccd607dfOr a specific algorithm:
$ avz fingerprint --algorithm sha256 employees.avro
3c9dd71e34662cb613aac0d4bdb7afa7309f2712ff97c1991a29028fccd607dfSupported: rabin (CRC-64-AVRO), md5, sha256, all (default).
Validate file integrity (reads every record):
$ avz validate employees.avro
Validated 8 records in employees.avro
employees.avro: OKCheck schema compatibility:
$ avz validate employees.avro --reader-schema new_schema.json
employees.avro: COMPATIBLESearches the JSON representation of each record and prints the entire matching record:
$ avz grep "ENGINEERING" employees.avro{"id":1,"name":"Alice Chen","department":"ENGINEERING","salary":125000.0,"email":"alice@example.com","tags":["rust","backend","senior"],"active":true}
{"id":3,"name":"Carol Davis","department":"ENGINEERING","salary":140000.0,"email":"carol@example.com","tags":["rust","systems","principal"],"active":true}
{"id":7,"name":"Grace Kim","department":"ENGINEERING","salary":135000.0,"email":"grace@example.com","tags":["frontend","react","senior"],"active":true}Pretty-print matches:
$ avz grep --pretty "rust" employees.avro{
"id": 1,
"name": "Alice Chen",
"department": "ENGINEERING",
"salary": 125000.0,
"email": "alice@example.com",
"tags": [
"rust",
"backend",
"senior"
],
"active": true
}
...Case-insensitive:
$ avz grep -i "alice" employees.avro
{"id":1,"name":"Alice Chen","department":"ENGINEERING","salary":125000.0,...}Fixed string (no regex — useful when pattern has special chars like *, ., ():
$ avz grep -F "125000.0" employees.avro
{"id":1,"name":"Alice Chen","department":"ENGINEERING","salary":125000.0,...}Count matches:
$ avz grep -c "ENGINEERING" employees.avro
3Invert match (show non-matching records):
$ avz grep -v -c "ENGINEERING" employees.avro
5| Flag | Description |
|---|---|
-i |
Case-insensitive matching |
-v |
Invert match (show records that do NOT match) |
-c |
Show only the count of matching records |
-F |
Treat pattern as a fixed string, not a regex |
--pretty |
Colorized pretty-print output |
Convert newline-delimited JSON to an Avro file:
$ avz fromjson --schema schema.json --output employees.avro employees.jsonl
Wrote 8 records to employees.avroUsage: avz fromjson [OPTIONS] --schema <SCHEMA> --output <OUTPUT> [INPUT]
Options:
-s, --schema <SCHEMA> Path to the Avro schema JSON file
-o, --output <OUTPUT> Output Avro file path
-c, --codec <CODEC> Compression codec [default: null]
[INPUT] Input JSON file (reads from stdin if omitted)
With compression:
$ avz fromjson --schema schema.json --output data.avro --codec snappy input.jsonlFrom stdin:
$ cat records.jsonl | avz fromjson --schema schema.json --output data.avroMerge multiple files into one:
$ avz concat employees.avro employees2.avro --output merged.avro
Concatenated 16 records from 2 files into merged.avroChange the compression codec of an existing Avro file:
$ avz recodec employees.avro --codec zstandard --output employees-zstd.avro
Re-encoded 8 records with codec 'zstandard' to employees-zstd.avroVerify the codec changed:
$ avz meta employees-zstd.avro | grep codec
avro.codec zstandardGenerate random records from a schema:
$ avz random --schema schema.json -n 3 --seed 42{"id":52656,"name":"Kate Kim","department":"ENGINEERING","salary":83049.87,"email":"carol821@example.com","tags":["mu xi lambda","iota"],"active":false}
{"id":5028,"name":"Kate Thomas","department":"MARKETING","salary":101244.93,"email":"bob447@test.org","tags":["delta lambda"],"active":true}
{"id":80872,"name":"Grace Smith","department":"ENGINEERING","salary":158813.58,"email":"hank875@example.com","tags":["pi nu"],"active":true}Pretty-print:
$ avz random --schema schema.json -n 2 --seed 42 --pretty{
"id": 52656,
"name": "Kate Kim",
"department": "ENGINEERING",
"salary": 83049.87,
"email": "carol821@example.com",
"tags": [
"mu xi lambda",
"iota"
],
"active": false
}Write directly to Avro format:
$ avz random --schema schema.json -n 1000 --format avro --output test.avro| Flag | Description |
|---|---|
-s, --schema |
Path to Avro schema JSON file (required) |
-n, --count |
Number of records to generate (default: 10) |
--seed |
Random seed for reproducible output |
-f, --format |
Output format: json (default) or avro |
-o, --output |
Output file path (required for avro format) |
--pretty |
Colorized pretty-print (json format only) |
All read commands work with S3 URIs. AWS credentials are loaded from the standard chain (env vars, ~/.aws/credentials, IAM role, etc.).
# single file
avz head -n 5 "s3://my-bucket/data/events.avro"
# glob pattern on S3
avz count "s3://my-bucket/data/dt=2026-03-16/*.avro"
# grep across S3 files
avz grep -F "transaction_id" "s3://my-bucket/data/*.avro"S3 files are downloaded into memory. For very large individual files, consider downloading first with
aws s3 cp.
| Codec | Flag value |
|---|---|
| None | null |
| Deflate | deflate |
| Snappy | snappy |
| Zstandard | zstandard or zstd |
| Bzip2 | bzip2 or bzip |
| XZ | xz |
MIT OR Apache-2.0