A fast, local CLI for inspecting CSV files and surfacing structural issues. Designed for operators who need answers immediately — no schemas, notebooks, or setup.

cargo install --path .

Or build from source:

cargo build --release
./target/release/csvops --help

# Inspect a CSV file
csvops profile data.csv

# Output as JSON for scripting
csvops profile data.csv --json

One command. Immediate answers.

• Delimiter detection — Reads first 8KB, scores consistency across comma, tab, pipe, semicolon
• Type inference — Classifies each column as integer, float, boolean, datetime, or string based on value patterns
• Missing value detection — Counts empties, NA, N/A, NULL, None, NaN; reports percentage per column
• Numeric statistics — Min, max, mean, standard deviation, percentiles (p25, p50, p75, p99) via streaming algorithms
• Cardinality estimation — Exact distinct counts up to 10K; HyperLogLog approximation above that
• Top values — Most frequent values per column, useful for spotting unexpected categories or data entry errors
• ID column detection — Flags columns that look like identifiers (high uniqueness, UUID patterns, naming conventions)
• Constant detection — Flags columns where every row has the same value
• Mixed-type warnings — Alerts when a column contains multiple data types (e.g., integers mixed with strings)
• Outlier flagging — Surfaces numeric values far outside the normal range

No configuration files. No database connections. No cloud accounts.

csvops profile data.csv

Options:

• --json — Output as JSON
• --delimiter <char> — Specify delimiter (auto-detected by default)
• --no-header — Treat first row as data
• --missing <tokens> — Custom missing value tokens (comma-separated)
• --sample-size <n> — Sample size for statistics (default: 10000)
• --no-color — Disable colored output

For files with a time column, surface how data changes across time periods:

csvops drift data.csv --time-col created_at --grain week

This groups rows by time bucket and highlights shifts in row counts, missing rates, and numeric distributions.

Options:

• --time-col <name> — Column containing timestamps (required)
• --grain <day|week|month> — Bucket granularity (default: day)
• --json — Output as JSON

File Statistics
──────────────────────────────────────────────────
  File:       sales.csv
  Size:       1.25 MB
  Rows:       50,000
  Columns:    8
  Delimiter:  ','
  Header:     yes

Warnings
──────────────────────────────────────────────────
  WARNING [revenue] Potential outliers detected
  INFO [customer_id] Appears to be an identifier column
  INFO [status] Low cardinality (3 distinct): may be categorical

Column Details
────────────────────────────────────────────────────────────────────────────────
  revenue
    Type:         float
    Missing:      120 (0.2%)
    Cardinality:  48,532
    Mean:         1,234.56
    Std Dev:      567.89
    Min:          0.01
    Max:          99,999.99

Delimiters: comma, tab, pipe, semicolon (auto-detected)

Date formats: ISO 8601, US (MM/DD/YYYY), Unix timestamps

Missing values: empty, NA, N/A, NULL, None, -, NaN (configurable)

• Streaming — handles large files with constant memory
• Reservoir sampling for percentile estimation
• HyperLogLog for high-cardinality columns
• Single-pass statistics

MIT