# Remove Legacy CLJS CLI ## Problem Logseq has one shipped CLI implementation under `cli/`, built from OCaml with Dune, Melange, and Vite, but it still retains most of the superseded Shadow ClojureScript implementation under `src/main/logseq/cli/` and its tests under `src/test/logseq/cli/`. The release migration deliberately removed the `:logseq-cli` Shadow build and made `cli/_build/default/dist/logseq-cli.js` the only producer of the staged `static/logseq-cli.js` artifact. `package.json`, CLI E2E preflight, npm package preparation, and Desktop runtime staging all use that OCaml/Melange artifact. `scripts/test-cli-release-config.mjs` rejects any active Shadow CLI build path. The remaining legacy command tree has no production entry point. Repository-wide namespace searches find no production consumer outside its own implementation. It contains 31 production files and about 14,800 lines, while 26 dedicated test files add about 16,800 more lines. Keeping this parallel implementation makes search results, tests, and maintenance work describe a binary that users cannot build or run. Not every namespace under `src/main/logseq/cli/` is obsolete. Electron, db-worker-node, and worker handlers still import shared daemon and transport helpers. Those imports are part of the active runtime and must remain. ## Proposal Delete only the legacy CLJS CLI closure that is unreachable from active production builds: - `src/main/logseq/cli/command/`; - `src/main/logseq/cli/auth.cljs`; - `src/main/logseq/cli/commands.cljs`; - `src/main/logseq/cli/completion_generator.cljs`; - `src/main/logseq/cli/config.cljs`; - `src/main/logseq/cli/format.cljs`; - `src/main/logseq/cli/humanize.cljs`; - `src/main/logseq/cli/main.cljs`; - `src/main/logseq/cli/output_mode.cljs`; - `src/main/logseq/cli/tree_text.cljs`; and - `src/main/logseq/cli/uuid_refs.cljs`. Delete their implementation-specific CLJS tests: - `src/test/logseq/cli/command/`; - `src/test/logseq/cli/auth_test.cljs`; - `src/test/logseq/cli/commands_test.cljs`; - `src/test/logseq/cli/completion_generator_test.cljs`; - `src/test/logseq/cli/config_test.cljs`; - `src/test/logseq/cli/format_test.cljs`; - `src/test/logseq/cli/main_test.cljs`; - `src/test/logseq/cli/output_mode_test.cljs`; and - `src/test/logseq/cli/uuid_refs_test.cljs`. Retain the active CLJS runtime boundary and its tests: - `logseq.cli.common` and `logseq.cli.common.db-worker`, used by Electron and worker handlers; - `logseq.cli.root-dir`, used by db-worker-node and worker logging; - `logseq.cli.server`, used by Electron runtime management; - `logseq.cli.transport`, used by Electron database access; - `logseq.cli.style`, used by db-worker-node; - `logseq.cli.log` and `logseq.cli.profile`, used transitively by the active server and transport helpers; and - `logseq.cli.test-helper`, which is also consumed by db-worker-node tests. Keep the shared helpers in their existing namespace for this decision. Moving them to a neutral namespace would add unrelated churn without reducing the duplicate CLI implementation. Update `docs/cli/logseq-cli.md` so its query example points to the shipped OCaml implementation in `cli/lib/query.ml` instead of `logseq.cli.command.query`. Preserve `static/logseq-cli.js` as a generated artifact of the existing OCaml/Melange staging flow; it is not part of the source deletion. ## Alternatives considered ### Retain the CLJS implementation as a reference Rejected because the OCaml implementation, its unit/parity tests, CLI E2E manifests, and the public command reference already own shipped behavior. A second non-buildable implementation is a stale source of truth rather than an executable reference. ### Delete every `logseq.cli` namespace Rejected because Electron, db-worker-node, and worker production code still consume the daemon, root-directory, transport, formatting, and shared database helpers listed in the proposal. ### Move the retained helpers before deleting the command tree Rejected for this decision because namespace relocation changes many active runtime imports but is not required to remove the parallel CLI. It can be considered separately if the `logseq.cli` name remains confusing after the obsolete implementation is gone. ## Acceptance criteria - The 31 unreachable CLJS source files and 26 implementation-specific test files listed by the proposal are removed. - The eight active CLJS production namespaces and their focused tests remain, and repository-wide namespace search still accounts for every Electron, db-worker-node, worker, and test-helper consumer. - `shadow-cljs.edn`, root build scripts, CLI E2E preflight, npm package preparation, and Desktop staging continue to identify `cli/_build/default/dist/logseq-cli.js` as the sole CLI producer. - `docs/cli/logseq-cli.md` no longer points readers to a deleted CLJS command namespace. - `node scripts/test-cli-release-config.mjs` passes. - `pnpm --dir cli test` passes. - `bb -f cli-e2e/bb.edn test --skip-build` passes against freshly staged CLI and db-worker-node artifacts. - Focused CLJS tests for `logseq.cli.common`, `logseq.cli.common.db-worker`, `logseq.cli.log`, `logseq.cli.profile`, `logseq.cli.root-dir`, `logseq.cli.server`, `logseq.cli.style`, and `logseq.cli.transport` pass. - `bb dev:lint-and-test` passes. ## Risks - A developer may have used the old CLJS namespaces manually even though they have no build, package, or documented public entry point. This decision intentionally gives up that unsupported workflow. - Removing duplicate CLJS tests reduces coverage of the deleted implementation. Shipped CLI behavior remains covered by `cli/test/` and `cli-e2e/`; active shared CLJS runtime tests remain in place. - The retained helpers continue to use the `logseq.cli` namespace, so the name will describe runtime infrastructure as well as the historical location of the removed CLI. Renaming them is outside this simplification. ## Questions - None.