📍 Roadmap: v0.4 (DataFrame I/O, CLI power, SQL, formats, parallelism)

[bytebrujo]

Tracking issue for the v0.4 roadmap from the June 2026 review. All P0–P3 gaps from docs/gaps.md (phases 1–4) are complete; this is the next wave.

High impact (phase-5)

• DataFrame write-back: toCSV / toJSON / save / write #29 DataFrame write-back (toCSV / toJSON / save / write)
• Expose DataFrame operations in the CLI (join, groupby, dedup, sample, transpose, pivot, split) #30 Expose DataFrame ops in CLI (join, groupby, dedup, sample, transpose, pivot, split)
• SQL query command: turbocsv query "SELECT ..." #31 SQL query command (turbocsv query)
• Activate parallel parsing to hit 500 MB/s throughput target #32 Activate parallel parsing (500 MB/s target)

Medium impact (phase-6 — formats & interop)

• More convert targets: Parquet, Arrow/IPC, Excel (.xlsx) #33 More convert targets: Parquet, Arrow/IPC, Excel
• turbocsv diff: key-aware diff between two CSV files #34 turbocsv diff (key-aware)
• turbocsv schema / infer: detect column types and export schema #35 turbocsv schema / infer
• Expose encoding selection/detection in the CLI #36 Expose encoding selection/detection in CLI

Polish (phase-7 — DX)

• Progress bars / ETA for large-file CLI operations #37 Progress bars / ETA
• Automatic resource cleanup via using / Symbol.dispose #38 Auto resource cleanup (using / Symbol.dispose)
• Worker-thread / Web Worker helper for parallelism #39 Worker-thread / Web Worker helper
• Remove stale empty file src/cli/commands/st.ts #40 Remove stale src/cli/commands/st.ts

Suggested order: start with #29 (self-contained, unblocks workflows), then #31 and #32.