nijaru/nu-data

Nushell as a Data Tool

Core Principle

Nu treats structured data as native tables. Reach for it over jq when exploring unknown shapes, doing multi-step transforms, or working with multiple formats.

When to Use vs. jq

| Situation | Tool | | -------------------------------- | ----------- | | Exploring unknown JSON structure | nu | | Multi-step filter + reshape | nu | | Multi-format (JSON + CSV + TOML) | nu | | One-liner in a bash/fish script | jq | | CI/CD pipeline (portability) | jq | | Pure regex text matching | grep/rg |

Common Patterns

Open and explore

open data.json                          # auto-parsed table
open data.json | explore               # interactive TUI browser
open data.json | describe             # show schema

Filter and select

open users.json | where age > 30
open users.json | get users | select name email
open users.json | get users.0.name    # dot-path indexing

Multiple files / formats

open data.csv | where status == "active" | to json
open config.toml | get database.host
ls *.json | each { |f| open $f.name | get id } | flatten

NDJSON (newline-delimited)

open logs.ndjson | lines | each { |line| $line | from json } | where level == "error"

Large data with Polars (requires plugin)

# Install once: plugin add ~/.cargo/bin/nu_plugin_polars
polars open large.parquet | polars filter col("age") > lit(30) | polars collect

Reshape and output

open data.json | get items | select name value | to csv
open data.json | update price { $in * 1.1 } | to json --indent 2

Quick Reference

| Task | Command | | ----------------- | ------------------------------------ | | Pretty print JSON | open f.json | | Count records | open f.json \| length | | Unique values | open f.json \| get field \| uniq | | Sort | open f.json \| sort-by field | | Group by | open f.json \| group-by field | | Flatten nested | open f.json \| flatten | | Convert format | open f.csv \| to json | | Filter nulls | open f.json \| where field != null |

Running from Fish

nu -c "open data.json | where active == true | to json"

Or pipe into nu:

cat data.json | nu -c "from json | where active == true"

Common Mistakes

• open auto-detects format by extension — for ambiguous files use open f.txt | from json
• Nu paths use . not [] for array index: get users.0 not get users[0]
• String interpolation uses $"...": $"Hello ($name)"
• Pipelines in each blocks need explicit params: { |row| ... } not { ... }