[renovate]

This PR contains the following updates:

Package	Change	Age	Confidence
@electric-sql/pglite (source)	0.3.14 → 0.5.11

---

Release Notes

electric-sql/pglite (@​electric-sql/pglite)

v0.5.11

Compare Source

Patch Changes

• 5573bb0: bug fix: check ready before close
• 2372f39: Update submodule
• 4321a41: Fix premature ReadyForQuery responses after extended-query errors, preventing PostgreSQL clients from desynchronizing before Sync.

v0.5.10

Compare Source

Patch Changes

• 367ee2f: Bug fix: wait for running queries to finish before doing a data dir dump
• cb06c21: Wait for queries to finish before closing PGlite instance. See #​1128
• 78012ca: Fix a windowed live.query (one given offset and limit) leaking its live_query_<id>_get_total_count prepared statement on unsubscribe. Only live_query_<id>_get was deallocated, so the count statement survived every teardown while its backing view was dropped, accumulating dangling prepared statements for the lifetime of the database.

v0.5.9

Compare Source

Patch Changes

• 389b731: upgrade to PostgreSQL 18.6
• 389b731: Export additional core symbols from the main WASM module required by the plpgsql_check extension

v0.5.8

Compare Source

Patch Changes

• c771db3: Do not set process.exitCode at all

v0.5.7

Compare Source

Patch Changes

• 2f9cf75: Fix convert_to

v0.5.6

Compare Source

Patch Changes

• 69b7d87: Apply the second options argument when calling PGlite.create(undefined, options).
• 6c14380: fixes for process.exitCode

v0.5.5

Compare Source

Patch Changes

• a290741: Reset retained protocol parser state after a malformed backend message so later queries can recover.
• 6b6f28d: Fix PGlite.create({ fs }) on a fresh database calling the provided filesystem's init() twice: the inner initdb instance no longer inherits the user-provided fs and always runs on its own scratch filesystem. Previously any VFS holding exclusive resources (e.g. OPFS sync access handles) failed with a contention error on first create.
• 354f4ae: Fix formatQuery (used by live.query/live.incrementalQuery to inline parameters) emitting %NL instead of the positional %N$L format specifier. A bare %NL is "min width N" and consumes format() arguments sequentially, so placeholders that are out of textual order silently bound the wrong values, and repeated placeholders failed with too few arguments for format().
• 219af1e: Expose the fields node-postgres derives from the CommandComplete command tag on Results: command (e.g. SELECT, INSERT, CREATE), and rowCount (the per-statement count from the tag). The tag was already received and parsed internally to compute affectedRows, but these values were not surfaced. Matching node-postgres' command/rowCount result fields lets pg-compatible adapters report them without re-parsing SQL.
• 20839a7: Preserve the host process.exitCode when closing a PGlite instance. close() calls _emscripten_force_exit(0), whose Emscripten runtime sets process.exitCode = 0, clobbering any exit code the host process had already set. This mirrors the existing save/restore guards in #init() and execProtocolRaw(), so closing a database no longer silently resets the host process's exit code.
• 7e784a4: Reject operations performed through transaction handles after the transaction has closed.

v0.5.4

Compare Source

Patch Changes

• 7e0d6d1: Detect the Node environment via process.type instead of process.versions.electron. The previous Electron guard (#​951) also treated the Electron main and utility processes as non-Node, which broke PGlite's filesystem code path there. process.type only excludes Electron's web contexts (renderer, web worker, service worker), so PGlite keeps using the Node.js path in the Electron main and utility processes. Follow-up to #​951 / #​813.

v0.5.3

Compare Source

Patch Changes

• 2ccbb4c: Improve parsing/raw results handling

v0.5.2

Compare Source

Patch Changes

• 21fc995: fix serializing arrays of BigInts
• 0720cb6: Alow serializing untyped parameters
• e09535f: add pg_stat_statements.tar.gz to main package
• a4e163a: fix process exit code set/restore

v0.5.1

Compare Source

Patch Changes

• 0641b37: Fix custom filesystems not receiving times set with FS.utime. Emscripten's FS.utime calls setattr with atime and mtime, but the BaseFilesystem bridge only handled the older timestamp field, so utimes was never called on OPFS AHP or any other BaseFilesystem. Files restored from a tarball kept the time they were written instead of the time stored in the tarball.
• e752dbe: Run initdb with TZ=UTC, like the backend, so a new database gets timezone = 'UTC' in postgresql.conf. Before, initdb wrote the host's zone as a fixed offset without daylight saving (e.g. Etc/GMT+5 for America/New_York), so timestamptz input without an offset was stored at a different instant on each non-UTC machine.
• c31130c: Start under the Node.js permission model (node --permission). Emscripten's NODEFS read the fs open flags with process.binding('constants'), which Node always denies in that mode, so new PGlite() threw ERR_ACCESS_DENIED even for memory://. The bundled build now reads them from require('fs').constants, which holds the same values.
• 0641b37: Fix files restored from a tarball getting a January 1970 modification time. loadTar passed the tar header's time to Emscripten's FS.utime in seconds, but FS.utime takes milliseconds. This affected every new database (initdb's data dir is restored from a tarball) and every loadDataDir. With the Node filesystem the files on disk got those dates too.

v0.5.0

Minor Changes

• 93d50aa: Upgrade to Postgres 18.3; move other extensions to their own npm packages;

v0.4.6

Patch Changes

• 2aa4d1a: Allow parsing of nulls in arrays #​997
• 2095d4e: Add ICU support.
• e937669: Restore process.exitCode
• 54ed6dc: new API method copyToFs() and new param postgresqlconf
• 817d073: restore exit code

v0.4.5

Compare Source

Patch Changes

• c6bddde: Fix caching of artifacts such that they are not downloaded multiple times

v0.4.4

Compare Source

Patch Changes

• b88c5c3: Disable checkpointer

v0.4.3

Compare Source

Patch Changes

• 2ae666f: Default database, user and role are now all "postgres"
• fb95e66: Allow setting initial memory size.
• 65fc101: Disable background workers.

v0.4.2

Compare Source

Patch Changes

• 41632c4: Allow passing initdb.wasm asset for bundlers that need it.

v0.4.1

Patch Changes

• 37fb39e: clear timers on exit; remove pglite-socket dependency on pglite-postgis

v0.4.0

Minor Changes

• d848955: New simplified PGlite with separate initdb.
  New included extension: pg_textsearch (experimental).
  New package for postgis (experimental) as extension.
  Breaking changes: 'postgres' is the default database instead of 'template1'.

v0.3.16

Compare Source

Patch Changes

• 3dfa40f: Add Apache AGE graph database extension support

v0.3.15

Compare Source

Patch Changes

• 45bff97: added pgcrypto extension
• 5ec474f: Added pg_hashids extension.

---

Configuration

📅 Schedule: (UTC)

• Branch creation
  • At any time (no schedule defined)
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR is behind base branch, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.