# Changelog

All notable changes to this project are documented here. The format is based on
[Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and this project aims to follow
[Semantic Versioning](https://semver.org/spec/v2.0.0.html).

## [0.3.0] - 2026-09-26

### Added
- **Versioning tab** on every table: a git-style graph of the snapshot DAG with one lane per
  Iceberg branch/tag, fork curves, ref pills, the current snapshot and an optional view of
  snapshots no ref reaches. Clickable counters open the branch, tag (with retention policy) and
  snapshot lists; the snapshot list (details + rollback) replaces the former Snapshots tab.
  Nessie catalogs also list their catalog-level references. Backed by a new
  `GET …/tables/{table}/versioning` endpoint.
- **Configurable dashboard**: DB-backed, live widgets (storage overview, maintenance
  reliability, hot partitions, file-size distribution, catalogs, executions…) that can be
  added from a gallery, pinned from a table's Storage/Overview tab, removed and reordered by
  drag and drop in edit mode (Flyway `V10`).
- **Storage & Health**: per-file data viewer (data + position/equality delete files) as a third
  drill-down level, "show all" files with type filter and sorting, typed per-key partition
  filters, multi-select partitions with a partition-scoped `rewrite_data_files` (Spark, live
  progress), a maintenance reliability card and "recently active partitions" (Nessie-aware).
- **Pipelines**: rerun a whole run or retry a failed task in place (resumes the skipped
  downstream tasks), per-task retries with a delay between attempts, and a graphical cron
  builder with a next-run preview.
- **Timeline**: clicking a Nessie snapshot fetches its real Iceberg summary on demand.
- **Public read-only demo mode**: `ICEGUARD_DEMO_MODE=true` on the frontend container (read at
  startup, served as `/config.js`, so it works with the published image) shows a "read-only demo"
  banner and refuses writes client-side with a clear message. `VITE_DEMO_MODE=true` does the same
  in local dev. Pair it with a server-side guard (e.g. a reverse proxy allowing only `GET` on `/api`).
- **Demo video voice-over** (`tooling/demo`, `npm run narrate`): English narration generated
  locally with Kokoro TTS, muxed onto the demo video with matching `.srt` captions.
- **Helm chart** (`charts/iceguard`) to deploy on Kubernetes: frontend + backend against a
  PostgreSQL you provide (the chart ships no database), pulling the published GHCR images.
  Ingress, network policy, OIDC, probes, PDBs and a `helm test` check are included, and the
  render fails fast when no database is configured. Installable without Helm too, via
  `helm template … | kubectl apply -f -`. CI lints and schema-validates the rendered manifests.
- **Configurable database schema** (`iceguard.db.schema`, env `ICEGUARD_DB_SCHEMA`; default
  `public`). IceGuard can now share a database with another application instead of owning
  `public`. Flyway creates the schema and Hibernate validates against it.
  **Upgrading:** the migrations were de-qualified (`public.x` → `x`) to make this possible, so
  their checksums changed. An existing database fails with `Migration checksum mismatch` until
  it is realigned once with `quarkus.flyway.repair-at-start=true` (checksums only, no data
  touched); drop the flag afterwards.

### Changed
- The "Lineage" tab is renamed **Evolution**; the standalone **Snapshots** tab moves into
  Versioning.
- The Nessie history call is hard-capped at 12 s, so a slow or hanging Nessie server falls
  back quickly instead of blocking for about a minute.
- Upgrade **Apache Iceberg 1.10.0 → 1.11.0** (backend, bundled Spark image and the default
  `--packages` coordinates); Parquet moves to **1.17.1** to match what Iceberg 1.11 pulls.
  `TableMetadataParser.read(FileIO, InputFile)` was removed upstream — the Nessie snapshot/
  partition-activity readers now use the `read(FileIO, String)` overload. Added `iceberg-orc`
  (+ `orc-core`): 1.11's `FormatModelRegistry` registers all generic format models eagerly, so
  even Parquet-only reads need the ORC classes on the classpath.
- Image build stages are pinned to `--platform=$BUILDPLATFORM`. Maven and npm produce
  architecture-neutral output, so a multi-arch build (`linux/amd64,linux/arm64`) now runs them
  once, natively, instead of once per target under QEMU.

### Fixed
- **Spark SQL literals are now escaped correctly** in maintenance `CALL`s. Quotes were doubled
  (`'O''Neil'`), which Spark reads as two adjacent literals (`ONeil`), and backslashes were not
  escaped, so a partition value ending in `\` could break out of the partition-scoped `WHERE`
  (reported by CodeQL). Backslashes and quotes are now backslash-escaped in both the UI and the
  backend, and the `WHERE` reaches Iceberg verbatim.
- **Frontend image now runs as a non-root user.** `nginx.conf.template` and the assets under
  `public/` were copied with the checkout's `0640` mode, so the entrypoint died on
  "default.conf.template: Permission denied" unless the container ran as root (or kept
  `DAC_OVERRIDE`). A `chmod -R a+rX` at build time makes the image satisfy the `restricted`
  Pod Security Standard, with only `NET_BIND_SERVICE` for port 80.

## [0.2.1] - 2026-07-03

### Added
- **Full `rewrite_data_files` options**: strategy, sort order, `where` and the options map, each
  opt-in, in a shared editor used by the table maintenance dialog and the pipeline task editor
  (filtered per engine; the Java executor now honours `target-file-size-bytes` and
  `min-input-files`).
- **Off-peak detection**: a "Commit activity · off-peak" chart on the table Overview, finding the
  quietest window from the snapshot timestamps (`GET …/commit-activity`, Nessie-aware).
- **Assisted pipeline editor** on a dedicated page (`/pipelines/new`, `/pipelines/:id/edit`):
  target → tasks → schedule → details.
- **Dedicated Edit Catalog page** exposing every field (Bearer/OAuth2 credentials, arbitrary
  properties, S3 storage modes); credentials are only sent when changed.
- `docs/NESSIE.md`, and the demo recorder/renderer (`tooling/demo`).

### Changed
- Explorer sidebar shows loaders while namespaces and tables load (slow catalogs).
- README demo GIF/MP4 now served from the v0.2.0 release assets.

## [0.2.0] - 2026-06-22

### Added
- **Real maintenance via the Iceberg Java API**: actual data-file compaction and orphan-file
  removal on small, append-only tables (hard size/file-count limits; refuses above them, pointing
  to Spark).
- **Delete-file maintenance** (Spark): `rewrite_position_delete_files` and
  `rewrite_equality_delete_files` actions for merge-on-read tables.
- **Local Spark tuning** in Settings — driver/executor memory, cores, instances and free-form
  confs, applied to `local[*]` maintenance runs (registered clusters keep their own config).
- Maintenance UI: in-dialog **result + logs** panel, and a **last-run** date on each action card.
- `docker-compose.images.yml` to run the published images without a local build, and `docs/CI.md`
  documenting the pipelines and security bots.

### Changed
- Upgrade **Apache Iceberg 1.7.1 → 1.10.0** (backend); the bundled Spark image moves to
  **Spark 3.5.8 + Iceberg 1.10.0** (fetched from the dlcdn CDN).
- Removed the fake "Nessie" REST-fixture bridge from the dev stack; the dev sandbox now ships a
  single, **real** Nessie Catalog Server.
- Split the oversized `TableService` / `TableDetail`; tuned **Dependabot** (grouped, monthly).
- README: dynamic **CI**, **release** and **GHCR** badges; documented the no-build images stack.

### Fixed
- Table detail page no longer crashes when a table fails to load — it shows a graceful error state.
- **CORS**: allow the nginx UI origin (`:8090`) in the Docker stacks, so catalog create/edit from
  the browser is no longer rejected.

### Security
- Added `SECURITY.md`, **CodeQL** and **Trivy** image scanning, and protected the `main` branch.
- Stopped leaking internal exception details to clients (opaque error reference IDs).
- Removed a stray `key` token file and ignored `key`/`*.key`.

## [0.1.0] - 2026-06-20

Initial release — an open-source web console for Apache Iceberg™ tables.

### Added
- **Multi-catalog** support: REST, Nessie, Polaris and **Unity Catalog** (auth: None / Bearer /
  OAuth2). The catalog **vendor is persisted** and drives vendor-specific behaviour such
  as Nessie commit-log history.
- **Browse**: namespaces & tables tree; create namespaces; create / drop / rename tables; insert
  sample rows; data preview.
- **Schema & metadata**: multi-column schema editor, properties editor and partition evolution,
  each applied in a single commit.
- **Storage**: point-in-time totals, file-size histogram, per-partition aggregates and file
  drill-down.
- **Lineage & history**: schema-version history with column diffs and snapshot-to-snapshot diff.
- **Maintenance**: expire snapshots, rewrite data files, rewrite manifests, remove orphan files,
  rollback — via a Java API executor (analyse) or a Spark executor (real `rewrite_data_files`).
- **Pipelines**: chain maintenance actions on a cron schedule with an Airflow-style run view.
- **Alerts**: threshold rules on table metrics with optional SMTP notifications.
- **Timeline**: snapshots + executions on one timeline.
- Per-table override of the Overview gauge thresholds (data files / snapshots).
- A real Nessie Catalog Server in the Docker dev sandbox.
- Published container images on GHCR (`iceguard-backend`, `iceguard-frontend`).

[0.3.0]: https://github.com/adelfardi/iceguard/compare/v0.2.1...v0.3.0
[0.2.1]: https://github.com/adelfardi/iceguard/compare/v0.2.0...v0.2.1
[0.2.0]: https://github.com/adelfardi/iceguard/compare/v0.1.0...v0.2.0
[0.1.0]: https://github.com/adelfardi/iceguard/releases/tag/v0.1.0