DataSquad Laptop Setup

A one-command macOS setup script that turns a fresh MacBook into a working data science consulting environment for the UCLA Library Data Science Center (DSC) and its student consultant program, DataSquad.

• Why this exists
• How it works
• Requirements
• Install
• Debugging
• What it sets up
• Known Issues
• Contributing
• License

Every academic term, DSC onboards new DataSquad student consultants and occasionally new staff, each of whom needs the same working set of tools before they can help a researcher clean a dataset, build a map, or debug an R script. Recreating that setup by hand from memory or an out-of-date wiki page is slow, inconsistent between machines, and silently drifts from what the job actually requires.

This script exists to make that setup reproducible and versioned: the toolchain is defined once, in code, and installing or catching up a machine is a single command instead of a checklist. It started as a fork of thoughtbot/laptop, a well-tested general web-dev setup script, and has been adapted over time for a data science consulting audience rather than a Rails shop.

As of July 2026, the tool list is no longer just inherited thoughtbot defaults — it's reviewed against what DSC consultations actually ask for. Real consultation-request data showed GIS work as the single largest consulting topic and Tableau as a frequently named tool, for example, neither of which had any supporting software installed until that review. See Known Issues for the full methodology and what's still pending a decision.

• Idempotent: safe to run repeatedly. It installs, upgrades, or skips each package based on what's already on the machine, so re-running it after a few months to catch a laptop up to the current tool list is normal usage, not a special "repair" mode.
• One source of truth: the entire package list lives in a single brew bundle block inside mac — no separate Brewfile to keep in sync, no tribal knowledge about what else to brew install by hand afterward.
• Tested on every change: a GitHub Actions workflow (.github/workflows/smoke.yml) runs the full script on a real, GitHub-hosted macOS runner for every push and PR, so a broken tap or a typo'd formula name fails CI instead of failing on a new hire's first day.
• Evidence over assumption: package additions and removals are expected to point at a reason — a documented DSC need, a verified dependency relationship (brew deps/brew uses), or real consultation-request volume — not just "this seemed useful." See Known Issues for the review trail.

Apple Silicon Macs running a current macOS release. The script has not been tested on Intel Macs or on macOS versions predating Apple Silicon, and no compatibility work is planned for those (DSC issues Apple Silicon MacBooks exclusively). Homebrew installs to /opt/homebrew on Apple Silicon rather than /usr/local; the script accounts for this directly rather than trying to detect architecture.

Download the script:

curl --remote-name https://raw.githubusercontent.com/UCLA-DataSquad/laptop/master/mac

Review the script (avoid running scripts you haven't read!):

less mac

Execute the downloaded script:

sh mac 2>&1 | tee ~/laptop.log

You'll need to use your laptop password. If it's the first time running this script, it will trigger an installation of Xcode and you will need to hit return.

Optionally, review the log:

less ~/laptop.log

Optionally, install thoughtbot/dotfiles.

It should take less than 15 minutes to install on a fresh machine (depends on your machine and network). Re-running it later to pick up new tools is much faster, since already-installed packages are skipped.

Your last run will be saved to ~/laptop.log. Read through it to see if you can debug the issue yourself. If not, copy the lines where the script failed into a new GitHub Issue for us. Or, attach the whole log file as an attachment.

macOS tools:

• Homebrew for managing operating system libraries.

Unix tools:

• Universal Ctags for indexing files for vim tab completion (built from HEAD; under review — see Known Issues)
• Git for version control
• OpenSSL for Transport Layer Security (TLS)
• RCM dotfiles management (thoughtbot template default; under review — see Known Issues)
• reattach-to-user-namespace tmux/pasteboard integration shim (thoughtbot template default; under review — see Known Issues)
• ripgrep fast recursive code/text search (replaces the older the_silver_searcher)
• Tmux for saving project state and switching between projects
• Watchman for watching for filesystem events (under review — see Known Issues)
• Zsh as your shell
• coreutils GNU command-line utilities for BSD/macOS parity (under review — see Known Issues)
• rlwrap readline wrapper
• pandoc markup converter (also used under the hood by Quarto)
• tree lists contents of directory in a tree like structure
• tealdeer fast tldr — short, example-driven command help for the CLI tools on this list
• jq JSON parser
• yq YAML/XML processor, same idea as jq -- used on project-registry.yaml, Carpentries lesson config.yaml, and GitHub workflow files
• wget network downloader
• rclone cloud storage data synch
• git-lfs Git support for large data files
• xan fast CSV toolkit for data work
• duckdb local SQL engine for querying/joining CSV, JSON, and Parquet files directly from the CLI, no server required
• docker container runtime (Docker Desktop, includes CLI + GUI)

• GitHub CLI for interacting with the GitHub API

Image tools:

• ImageMagick for cropping and resizing images

Geospatial:

• QGIS desktop GIS application — added because GIS is DSC's single largest consultation topic and QGIS is explicitly named in patron requests
• GDAL (ogr2ogr, gdalinfo) — CLI companion to QGIS for scripted format conversion and reprojection

Apps:

• OpenRefine for cleaning and transforming messy data
• iTerm2 terminal emulator
• Positron Posit's data science IDE (R/Python/Jupyter/Quarto in one editor); installed alongside RStudio and VSCodium, not a replacement for either

Languages and editors:

• R - Base R installed via Homebrew
• RStudio - R IDE
• Miniforge - conda-forge-based Python/conda distribution (follows The Carpentries' 2025 setup recommendation to move off Anaconda)
• pixi - fast, project-based Python/conda package manager, taught in DataSquad workshops
• uv - fast Python package/project manager (pip/venv/poetry replacement); installed so students and staff can drop into whatever setup a researcher's own project already expects (pyproject.toml, uv.lock, requirements.txt)
• VSCodium - Python/shell/git editor (VS Code without Microsoft branding/telemetry), also per The Carpentries' 2025 recommendation
• Node.js - pinned to the Active LTS line (node@24), for working in the Astro/Jekyll/Hugo site repos across DSC (Astro/Jekyll/Hugo themselves are scaffolded per-project via npm/bundler, not installed globally). Unversioned node tracks Homebrew's Current release, which changes out from under you and isn't recommended for production.
• Quarto - scientific and technical publishing system used for DSC reports, slides, and stats

Linting / repo maintenance:

• ShellCheck for shell script linting (used on mac itself, see Contributing below)
• yamllint for YAML linting

OSPO / RSE tooling — DataSquad students increasingly work as OSPO assistants on the software-curation side of open source program office work (license compliance, citation/publishing, security posture), so these are installed alongside the general toolchain rather than assumed to be looked up later:

• pre-commit manages multi-language git hooks (formatting, linting, secret scanning) from a single .pre-commit-config.yaml per repo
• Copier scaffolds a new repo from a template (tests, CI, docs, licensing, community files) and can reapply template updates to existing repos later
• REUSE SPDX-based license/copyright compliance; reuse lint catches missing or inconsistent license headers
• OSV-Scanner scans a repo's lockfiles/manifests against the OSV vulnerability database
• zizmor static analysis for GitHub Actions/Dependabot workflow configs; catches insecure CI setups (overly broad permissions, unpinned actions)
• OpenSSF Scorecard runnable locally, not just as a GitHub Action; checks a repo's CI practices, code review requirements, pinned dependencies, and other security posture signals

More specialized OSPO/RSE tooling (repository templates built with the above, OpenSSF Scorecard as a CI check, howfairis/FAIR assessment, ScanCode Toolkit for auditing inherited codebases, CodeMeta, workflow engines like Snakemake/Nextflow, portfolio-health metrics via CHAOSS) is documented separately in DataSquad's own OSPO-assistant documentation rather than installed universally here — not everything on that list belongs on every laptop.

See the wiki for more customization examples.

This script is a fork of thoughtbot/laptop (a general web-dev setup script), adapted over time for data science consulting work.

A July 2026 evidence-based review (validation-prompt-tooling-2026-07-26.md, cross-checked against DSC's own consultation-request data and verified against primary sources — Homebrew, Node.js, Astro, and Positron docs — before adopting) confirmed and removed several genuinely redundant/unused packages (gcc, xz, cairo, libxt, libyaml — either transitive dependencies of r already, or depended on by nothing installed here) and added tools backed by real demand (qgis, gdal, tableau-adjacent GIS/viz gap; duckdb, ripgrep, tealdeer). One caveat from that review: an external keyword-frequency claim that R now outranks Python in consultation requests could not be reproduced from the underlying data (the merged consultation dataset blends real patron intake with Trello project-card text duplicated once per comment, which distorts any raw frequency count) — treat any future frequency claim from this dataset with that in mind.

Still marked "under review," pending a decision (not yet confirmed either way): rcm, reattach-to-user-namespace, universal-ctags HEAD build, vim, watchman, coreutils. Also open: whether to add tableau-public (real demand, but Tableau Public auto-publishes saved workbooks to a public gallery by default — needs a documented opt-in rather than a silent install given patron-data handling) and whether to pilot jamovi as a gentler stats-GUI on-ramp for students.

1. Edit the mac file.
2. Document in the README.md file — say why, not just what, if you're adding or removing a package (see Why this exists).
3. Follow shell style guidelines by using ShellCheck and Syntastic. shellcheck is installed by mac itself (see above), so if you've already run the script once you have it:

shellcheck -s sh mac

4. Check out the GitHub Actions tab to make sure the build runs without error.

Thank you, contributors!

By participating in this project, you agree to abide by the thoughtbot code of conduct.

This repository is a fork of thoughtbot/laptop and retains thoughtbot's original MIT license, per the terms of that license (copyright 2011-2020 thoughtbot, inc.) — see the LICENSE file. UCLA DSC's changes on top of the original are contributed under the same license.