Every application ships with rows it needs to run. Roles and permissions. Plan tiers. Workflow states. Country and currency lists. Email templates. They change with each release, and customers’ admins edit them too. Then comes the upgrade, and somebody’s careful edit gets wiped, or the change the release was supposed to make never lands.
Bootstrapping keeps that reference data in step with the database. It loads the rows on first install and updates them on every release after, without undoing what admins changed. When the release and an admin changed the same value in different ways, it stops and asks a person to decide.
The Pitch in Four Numbers
- 492,228 values checked across 1,200 random histories of seed versions and admin edits, with none overwritten without a conflict.
- 400 applies killed partway through, and every database found exactly as before or exactly as after.
- 1 of 8 processes applying the same seed to one file at once actually wrote, in each of 60 rounds.
- 129,632 checks that values compare the way the database stores them, across 16 column types.
How It Works
Most teams handle this with a hand-written migration script per release. Those scripts either overwrite customer edits or miss changes the release meant to make. Upserting from a seed file doesn’t help, because it can’t tell an admin’s edit from a stale value, so it overwrites both.
Bootstrapping’s fix is memory. The database records the seed that was last applied, so every value can be compared three ways: the new seed, the live row and the last-applied seed. A change only in the seed is applied. A change only by the admin is kept. A value changed differently on both sides stops the apply and goes to a person. The homepage shows it on an editor role: a description changed only in the seed gets updated, a name changed only locally stays.
Seeds are TOML files, one per table, that go through code review like anything else. Each column can be owned by the merge, by the seed or by the local admin. Rows dropped from the seed can be kept, deleted or retired. Renames keep the row’s ID. Links between tables are written by name, long lists can live in CSV, and a link can match a row on several columns.
It works with SQLite, PostgreSQL, MySQL and MariaDB, with the same seeds and merge logic on all four. Values are compared as the database stores them, so 7 in a numeric(10,2) column counts as 7.00, and a value too long for its column stops the plan instead of being cut short. The record of what was applied lives in four small tables in the target database, written in the same transaction as the data.
How You Use It
As a Go library with any database/sql driver, with seeds embedded in the application and applied at startup. As a command-line tool, bootstrapping or bst, that plans, applies, explains individual values and lists history, with JSON output and exit codes for deploy scripts. And as a GitHub Action that posts the plan on a pull request and fails the check on a conflict.
Who It’s For
- SaaS and self-hosted software vendors shipping roles, permissions and plans that customers customise.
- Teams tired of per-release data migration scripts.
- Apps with editable templates and workflows, like helpdesks and CRMs, where admin changes are the whole point.
- Anyone shipping lookup lists, countries, currencies and categories, that need to stay current.
Limits
It fills tables and doesn’t create them. Schema migrations stay with another tool, and Bootstrapping runs after it. It’s built for reference data in the thousands of rows, not millions, because every managed table is read in full on each plan. And it supports those four databases only.
The Four-Question Screening
Every project featured here gets the same four questions, the way you’d screen a startup idea.
1. Where is the money in it?
Bootstrapping is open source under Apache 2.0, with no paid tier. Its buyers are software vendors whose customers run their own installs or heavily customise a shared one, and every bad upgrade costs those vendors support time and goodwill. Paid paths include support contracts, a hosted dashboard that shows seed state and conflicts across all customer installs, and integrations with the migration tools teams already use. Fleet-wide visibility is the part a vendor would pay for.
2. What real problem does it solve?
Upgrades that trample customer settings. It’s a small, recurring source of support tickets and broken trust: the admin spends an afternoon setting up roles, and the next release resets them.
3. The technology in plain words
Think of a shared recipe book. The publisher sends out new editions, and you’ve written notes in the margins. A careless update would print a fresh book and throw away your notes. Bootstrapping compares the old edition, the new edition and your copy line by line. It takes the publisher’s changes, keeps your notes, and when you both changed the same line, it asks you which to keep.
4. Moat and differentiation
Schema migration tools are plentiful. Tools for the data inside the schema, with a real three-way merge, are rare. The memory of the last-applied seed is the key idea, and the testing, half a million values, killed applies and racing processes, is what makes it trustworthy. The moat that can grow is support across databases and the review workflow on pull requests, which puts it in the release process.
Why It Matters
Customers judge software partly by whether upgrades respect their work. Getting seed data right is unglamorous, and it’s exactly the kind of thing that decides whether a customer trusts the next release.
Follow the Project
Read how the three-way merge works, start with the quick start, see how it was tested, or get the code from the Bootstrapping repository on GitHub. Bootstrapping is listed with the other open-source projects on SoftDevStudio.com, and more featured projects are in the Projects section.