Exporting and Batch Mode

VisiData is not just an interactive TUI — it is also a data conversion engine that can run headlessly in batch mode for scripted pipelines, and accepts piped input from any shell command.

Learning Focus

Use Ctrl+S to save from within the TUI. Use -b -o for headless batch conversion. Replay recorded CommandLog sessions for reproducible ETL pipelines.

In-App Saving

Save your current sheet without leaving VisiData:

Ctrl+S          save current sheet to file (prompts for path)
g Ctrl+S        save all open sheets
z^S             save current column only
Ctrl+D          save CommandLog to .vdj file

The save format is auto-detected from the file extension (.csv, .json, .tsv, etc.):

# Inside VisiData, after filtering/analysis:
Ctrl+S
# Enter path: /tmp/clean_data.csv

Combine with the filter workflow:

# 1. Filter rows with | regex or z| expr
# 2. Press " to open filtered sheet
# 3. Press Ctrl+S → enter filename
# 4. Enter to save

Stdin Pipe Input

Any command output can be piped into VisiData:

command | vd -f format

# Examples:
ps aux | vd -f fixed --skip 1
df -h | vd -f fixed --skip 1
cat access.log | vd -f txt
curl -s https://example.com/data.csv | vd -f csv
curl -s https://api.github.com/repos/saulpw/visidata/tags | vd -f json

Stdout Pipe Output

VisiData can output a sheet to stdout for further piping:

vd -b input.csv -o -           # output to stdout as TSV
vd -b input.csv -o - | grep "error"

Batch Mode: Data Conversion

Batch mode (-b) runs VisiData without the TUI:

# Convert CSV to JSON
vd -b input.csv -o output.json

# Convert JSONL to CSV
vd -b events.jsonl -o events.csv

# Convert Excel to TSV
vd -b report.xlsx -o report.tsv

# Convert SQLite table to CSV
vd -b site.sqlite -o tables_export.csv

# Convert fixed-width to JSON
ps aux | vd -f fixed --skip 1 -b -o processes.json

CommandLog Replay

VisiData records every action in a CommandLog (.vdj file). Save and replay sessions for reproducible batch processing.

# Inside VisiData:
Shift+D         # view CommandLog sheet
Ctrl+S          # save it as a .vdj file

Basic Replay

vd --play session.vdj

Parameterised Replay with field=value

Use {field} placeholders in your CommandLog to make it reusable:

# A CommandLog session.vdj contains:
# open-file  filename={input_file}
# save-sheet  filename={output_file}

# Replay with specific files:
vd --play session.vdj input_file=jan.csv output_file=/tmp/jan_clean.csv
vd --play session.vdj input_file=feb.csv output_file=/tmp/feb_clean.csv

This turns a recorded session into a reusable ETL script:

#!/bin/bash
for month in jan feb mar; do
  vd --play clean_pipeline.vdj \
    input_file=data_${month}.csv \
    output_file=/tmp/reports/${month}_clean.csv
done

-i — Interactive Mode After Replay

# Replay session, then open VisiData interactively
vd --play session.vdj -i

Useful for debugging: run replay → pause at last state → inspect manually.

Practical Pipelines

Nginx Log Analysis

# Extract and export
vd -b access.log -o /tmp/access_analysis.csv

Multi-File Batch Conversion

for f in ~/github/practice-folder/visidata/04-cleaning/*.csv; do
  base=$(basename "$f" .csv)
  vd -b "$f" -o "/tmp/json/${base}.json"
done

Daily Report Automation

DATE=$(date +%Y%m%d)
vd -b ~/github/practice-folder/visidata/05-logs/01-access.log \
   -o /var/www/html/reports/access_${DATE}.csv

Batch Mode Options

Option	Description
-b / --batch	Run without TUI
-o file / --output file	Save final sheet to file
--play file.vdj	Replay a saved CommandLog
--replay-wait N	Wait N seconds between replayed commands
-i / --interactive	Launch interactively after batch replay
field=value	Replace {field} placeholders in CommandLog

Troubleshooting

Problem	Cause	Fix
Batch outputs empty file	Pipe closes before read	Use explicit file instead of stdin
CommandLog replay fails	Sheet structure changed	Re-record the session
Format not detected in batch	No extension or stdin	Always use -f format in scripts
Encoding error in batch	Non-UTF-8 input	Add --encoding latin1

Best Practices

• Always specify -f in batch scripts — never rely on auto-detection in automation.
• Use CommandLog replay (--play) to document and share data transformation steps.
• Pipe vd -b input -o - into other tools (grep, jq, awk) for lightweight ETL.
• Set --max-rows when sampling large files in batch to avoid long runtimes.

What's Next

• Performance and Profiling — large files, sampling, lazy loading
• Opening Files and Formats — load CSV, JSON, SQLite, Excel, and more