v0.1.6

What's Changed

• On-disk representations of the expected schema show full function definitions for easier diffing
• Faster and more atomic reading from and writing schema to disk
• We now run some tests on MacOS in CI

Full Changelog: v0.1.5...v0.1.6

Note for MacOS: MacOS complains the bundle might be corrupted. As per Internet sources, this could be because the bundle is not signed (I don't have MacOS and don't pay for an Apple Developer license). I'm interested in hearing if it doesn't work for you.

v0.1.5

What's Changed

• Codd now provides an app bundle in Github releases for aarch64 macOS
• Codd now includes extended statistics and index validity in representations
• Defensively disregard non-persistent relations to avoid race conditions in postgres when verifying schemas while temporary relations are dropped by the application. For details see https://www.postgresql.org/message-id/flat/CACgY3QYe3mqMkODOzwx%3DEN4iGoddObUotLL8hGnCg%3D%3D3vmxd_g%40mail.gmail.com
• CI improvements, including automatically publishing binaries in Github releases from a Github workflow

Full Changelog: v0.1.4...v0.1.5

Note for MacOS: MacOS complains the bundle might be corrupted. As per Internet sources, this could be because the bundle is not signed (I don't have MacOS and don't pay for an Apple Developer license). I'm interested in hearing if it doesn't work for you.

v0.1.4

Major changes include:

• Show and store migration application time in codd_schema.sql_migrations
• Codd can now retry failed no-txn migrations from the last failed statement
• Accept whitespace in KVP connection strings just like psql
• Pg dumps that contain codd_schema are allowed as migrations in most cases, meaning it's easier to consolidate a lot of migrations into a single dump for faster database creation

v0.1.3

Codd no longer opens all pending migrations simultaneously. It opens one migration file at a time, ensuring it never runs into ulimit or system open file limits, which are particularly low in macOS. This means codd can now apply any number of migrations of any size, since not only the number of simultaneously open files is constant and small, but also memory usage is constant since migrations are streamed from disk.

v0.1.2

User visible changes:

• Officially support postgres 16
• Properly detect database settings, so SET can be used without affecting on-disk representations
• Disk representations are now pretty printed instead of being single line json
• SQL parser now assumes standard_conforming_strings is on. It might not parse SQL properly when it's off
• Support nested C-style comments and statements where semicolons are not terminators, such as CREATE RULE
• Better error messages for codd add
• Varchar columns' lengths are now part of on-disk representations
• The database role's search_path setting no longer interferes in representations
• Other smaller improvements in representations
• Fix building in macos
• All installation methods now point to the latest stable release, not the latest master

Internal changes:

• New darwin pipeline in CI that ensures installation works
• CI pipeline now populates cachix after builds
• Internal refactor to use type system to guarantee certain parts of the codebase run inside transactions and others don't

v0.1.1

Support for postgres 15, fixes for range types' constructors ownership and privileges, and changes in representation of privileges to better accomodate multiple grantors per grantee in the same database object.

v0.1.0

This release contains the self-contained x86_64-linux executable for codd 0.1.0