Read tasks the way Logseq does in every command (#92, part 1)

AGENTS.md

@@ -163,9 +163,12 @@ U=$(logseq-cli find-block --content "tag support" --page "Project Alpha" --exact
 # All open tasks
 logseq-cli get-todos
 
-# Filter by status
+# Filter by marker
 logseq-cli get-todos --status TODO --status DOING
 
+# Filter by state: open (the default), done or cancelled
+logseq-cli get-todos --state done
+
 # Filter by page (substring)
 logseq-cli get-todos --page "Project Alpha"
 
@@ -194,6 +197,14 @@ logseq-cli set-todo-status --id JOURNAL-REF-UUID --status DONE --follow-refs
 at the end; `followed` in `--json` lists the blocks it passed. A chain that
 loops or reaches a missing block is refused and nothing is written.
 
+`set-todo-status` changes a task only. A block Logseq reads no marker in
+(`todo x`, `TODO:`, `TODO` followed directly by a line break) is refused with
+`not_a_task`, and nothing is written; so is a block whose marker is not at the
+start of the text, and a block that only points to another one (the message
+names `--follow-refs`). `--content` that matches blocks but no task is refused
+with `no_task_matches`, unless `--follow-refs` is given. A heading task
+(`## TODO x`) keeps its heading: `## DONE x`.
+
 Do not use `replace-text` to change a marker: it rewrites by text match, so it
 also hits the word elsewhere on the page and silently retypes the rest of the
 line. `set-todo-status` swaps only the marker, in one call.
@@ -257,9 +268,15 @@ logseq-cli add-journal-content \
   --content "- ## Section\n\t- Child item\n\t\t- Grandchild"
 ```
 
-Each line is a block, with two exceptions that stay with the block they belong
+Each line is a block, with four exceptions that stay with the block they belong
 to, as in Logseq's own files. A `key:: value` line without a bullet goes on the
-block above. So does a code block: from a line starting with ```` ``` ```` or
+block above. So does a `SCHEDULED: <…>` or `DEADLINE: <…>` line, indented or
+not: `TODO x` with `SCHEDULED: <2026-09-25 Fri>` under it is one block, a task
+with a date. So does a `:LOGBOOK:` drawer, every line from `:LOGBOOK:` to
+`:END:`. A `:LOGBOOK:` that no `:END:` closes before the next bullet line is no
+drawer but text, as a code fence nothing closes is, and the lines after it are
+read as usual. So does a
+code block: from a line starting with ```` ``` ```` or
 `~~~` to the next such line without a bullet, every line is code, `- ` lines
 included, and keeps its indentation. Put the fence on a bullet line to make the
 code block a block of its own:
@@ -270,6 +287,12 @@ logseq-cli add-note-content --page "Notes" \
 # -> "Example" with the code on it, then a block holding only the sh code
 ```
 
+```bash
+logseq-cli add-note-content --page "Notes" \
+  --content $'TODO Send the report\nSCHEDULED: <2026-09-25 Fri>'
+# -> one task block with its date, not a task and a block that says SCHEDULED
+```
+
 Text the CLI writes as ONE block must come back from the page file as that
 block. Logseq's file parser takes a `- ` or `#` line after the first (indented
 too) and a code fence nothing closes as block boundaries, and rebuilds the page
@@ -440,12 +463,17 @@ rollback), and the message ends with how many, or "Nothing was written.".
 | `logseq_error` | Logseq threw on the write and answered with an error object | `method`, `logseq_message` |
 | `page_exists` | `create-page` on a page that exists, whose properties Logseq would drop | `page` |
 | `rename_refused` | `rename-page` onto a name another page has (Logseq would merge the two) or an empty one | `old`, `new`, `why` (`exists` or `empty`) |
+| `block_changed` | `set-todo-status`, `replace-text`, `set-property` and `remove-property` (every write that replaces a block's text) read the block again right before writing, and it no longer holds the text they read: someone changed it in between, and the write would have replaced that change. Nothing was written | `block` |
 | `write_not_verified` | The write does not show in Logseq: not written, or not all of it | `method`, `target`, `expected`, `got` |
 | `read_only` | Writes are switched off by `[safety] read_only`, `LOGSEQ_CLI_READ_ONLY` or `--read-only`; every command that writes refuses, `--dry-run` included, before its first request | `source` (list of `config`, `env`, `flag`), `config_path` (`null` when no config file was found) |
 | `config_error` | The config file cannot be used: it does not parse, `LOGSEQ_CLI_CONFIG` names a file that is not there, or `[safety]` does not check out: a key it does not know, a `[safety]` key outside `[safety]`, a value that is not a boolean, or a `[graph] keep_empty_blocks_last` that is not. A command that writes refuses rather than run without the limits the file may hold. Nothing was sent | none |
-| `block_not_found` | The block the command names does not exist; the placeholder Logseq keeps for a missing ref target counts as none | `id` |
+| `block_not_found` | The block the command names does not exist; the placeholder Logseq keeps for a missing ref target counts as none | `id`; for `set-todo-status --content` the `content` and `page` searched instead, no `id` |
 | `dead_ref` | `set-todo-status --follow-refs`: the chain of refs leads to a block that does not exist | `id` (the missing block), `followed` |
 | `ref_cycle` | `set-todo-status --follow-refs`: the chain of refs comes back to a block it passed | `id` (the block met again), `followed` |
+| `not_a_task` | `set-todo-status` on a block Logseq reads no marker in, or whose marker is not at the start of the text (after an optional `#` prefix and a space), or that only points to another block and `--follow-refs` is missing | `id`; `points_to` for the pointer case |
+| `no_task_matches` | `set-todo-status --content`: blocks match, none is a task; not given with `--follow-refs` | `content`, `page`, `match_count` |
+| `ambiguous` | `set-todo-status --content`: several blocks match | `content`, `page`, `matches` (the uuids of all candidates) |
+| `missing_selector` | `set-todo-status` without `--id` and without `--content` plus `--page`; refused before any request | none |
 
 `set-property` and `remove-property --name` check the page themselves and
 name `page` and `property` instead.
@@ -479,6 +507,11 @@ write into X). A `--keep-ids` write is refused while any block is open.
 `write_not_verified` can be a false alarm when someone else wrote the block
 between the write and the read-back; read the block before retrying.
 
+On `block_changed`: nothing was written to that block, and the change made
+elsewhere stands. Read the block again and repeat the command if it still
+makes sense. The check narrows the time between reading and writing to one
+request; it does not close it.
+
 `replace-text` writes the other blocks when one is refused, and names each
 refused block's reason in `failed_reasons` (`{id: reason}`) beside `failed`.
 
@@ -512,7 +545,7 @@ Every command with all its options: [docs/commands.md](docs/commands.md).
 | `insert-block` | Insert at specific position (after/before/child-of, `--first` for first child); `--keep-ids` preserves `id::` values in `--tree` or `--content` (also on `add-note-content`, `add-journal-block`, `add-journal-content`) |
 | `find-block` | Find blocks by content; `--limit N` caps the output (what is withheld goes to stderr); `--with-children` prints the subtree; `--uuid-only` prints bare uuids and fails on no match; `--exactly-one` fails unless exactly one block matches |
 | `update-block` | Change one block's content (by `--id` or `--where-content`); its properties are kept, a key `--content` sets takes the new value |
-| `set-todo-status` | Change a TODO/DOING/DONE marker (never `replace-text`) |
+| `set-todo-status` | Change a task's marker (never `replace-text`) |
 | `move-block` | Relocate a block, keeping its UUID and refs |
 | `copy-block` | Copy a block to another page (new UUID) |
 | `remove-block` | Delete a block by UUID |

CHANGELOG.md

@@ -12,6 +12,10 @@ file carries most of the reasoning behind the tool.
 
 ### Added
 
+- `get-todos --state open|done|cancelled` (repeatable, default `open`) selects
+  tasks by state, so `--state done` lists the finished ones and `--state
+  cancelled` both spellings of the cancelled marker (#92).
+
 - `[graph] keep_empty_blocks_last = true` (or `LOGSEQ_CLI_KEEP_EMPTY_BLOCKS_LAST`)
   makes a write that ends a section go before the empty blocks that end it, so
   they stay last and no empty line is left in front of the entry (#110). Found by
@@ -92,6 +96,80 @@ file carries most of the reasoning behind the tool.
 
 ### Changed
 
+- `analyze-journal-patterns` counts tasks from the database, one count per
+  journal day. `tasks` holds `open`, `done`, `cancelled` and `done_rate`, and
+  the `tasks` of each entry holds `open`, `done` and `cancelled`. A task is a
+  block that carries a marker Logseq reads, so `TODO:`, a lowercase `todo`,
+  `TODO` followed by a line break and Markdown checkboxes (`- [ ]`) are no
+  longer counted as tasks, and CANCELED and CANCELLED no longer count as done.
+  The rate is named `done_rate`, not `completion_rate`, on purpose: its value
+  changes (cancelled tasks do not count, checkbox lines drop out), and a key
+  must not keep its name while it returns a different number, so a reader of the
+  old key now fails loudly instead of reading a wrong figure. `done_rate` is
+  `done / (open + done)` in percent with one decimal. Tasks are dated by the
+  page their block is on. `get-todos --from/--to` also counts the days a block
+  ref carried a task to. Habit tracking still counts the text pattern `- [ ]` /
+  `- [x]` and now says so in `--help` and in the text output (#92).
+
+- `analyze-graph` counts tasks from the database, like `get-todos`:
+  `total_todos` is replaced by `tasks` (`open`, `done`, `cancelled`,
+  `open_by_marker`) and the text line reads `Tasks: N open (…), M done, K
+  cancelled`. Before, a pattern over the page text counted `TODO:`, a lowercase
+  `todo`, a block that starts with `[ ]` and a marker on a later line, so the
+  two commands disagreed on the same graph. A note says how many blocks start
+  with `[ ]`, which Logseq shows as text. `--days` limits only the Recently
+  updated list; its help now says so (#92).
+
+- `smart-query --request tasks` now also finds WAIT, WAITING, IN-PROGRESS and
+  STARTED tasks. The pattern is matched by every keyword it lists (`todo`,
+  `task`, `tasks`, `incomplete`, `pending`, `aufgaben`, `offene`, `offen`), not
+  only `tasks`, so all of them return more. The old list was TODO, LATER, NOW,
+  DOING; `--advanced` with your own query keeps it. `done` is unchanged (#92).
+
+- `set-todo-status --status` no longer tells upper case from lower case
+  (`--status done` is taken, as `--status todo` always was for `get-todos`),
+  and sets every marker Logseq's frontend knows: WAIT, WAITING, IN-PROGRESS and
+  CANCELLED are new. STARTED stays out: Logseq's parser reads it, but the
+  frontend does not offer it, so the CLI does not write a marker Logseq never
+  writes (#92).
+
+- `get-todos` without `--status` now lists every open task, not only TODO,
+  DOING, NOW and LATER: tasks marked WAIT, WAITING, IN-PROGRESS or STARTED come
+  with it. Logseq draws the first three as open tasks, and its parser reads
+  STARTED as a marker too, although its own interface does not offer it (#92).
+  The result is the same shape and every task carries its `marker`; only more
+  tasks appear. To keep the old selection, pass `--status TODO --status DOING
+  --status NOW --status LATER`. `--status` takes only the eleven markers Logseq
+  reads and does not mind the case (`--status todo` still works);
+  `get-todos --status foo` used to run and find nothing without a word, and is
+  now a usage error with a non-zero exit. `--status` has no default any more,
+  so `--state` decides unless `--status` is given, and a note says so when both
+  are. `--include-done` still works, adds DONE to either, and is left out of
+  `--help`. Tasks are ordered by marker, with IN-PROGRESS and STARTED among the
+  work in hand and WAIT and WAITING after LATER; they used to sort after DONE,
+  by page.
+
+- `set-todo-status` refuses a block that is no task instead of making it one:
+  `no marker` became `DONE no marker`, `todo x` became `DONE x`. Logseq reads
+  no marker in such a block (`todo x`, `TODO:`, `TODO` followed directly by a
+  line break); the command now stops with `reason: "not_a_task"` and writes
+  nothing, also under `--dry-run`. A block that only points to another one is
+  refused too, naming `--follow-refs`. With `--content`, blocks that match but
+  none of which is a task are refused (`no_task_matches`) instead of one being
+  rewritten; with `--follow-refs` this holds for the block at the end of the
+  chain, and a match that only points to a task is followed as before. Under
+  `--json`, a block not found by `--content`, an ambiguous match and a missing
+  selector now answer with a JSON object with a `reason` (`block_not_found`,
+  `ambiguous`, `missing_selector`) instead of plain text. Without `--json`,
+  these three messages now start with `Error: `, like every other refusal; the
+  wording after it is unchanged (#92).
+
+- `get-todos` (`content`, `--match`, `--tag`) now reads an unclosed
+  `:LOGBOOK:` as text and a `:LOGBOOK:` inside a code fence as code, the way
+  mldoc reads them (measured with mldoc 1.5.7); before, everything after such
+  a line was dropped from the task text. A `SCHEDULED:` or `DEADLINE:` line inside a code
+  fence and a stray `:END:` now stay too (#92).
+
 - `doctor` no longer advises removing a config file that does not parse:
   "Fix the config file." Once the file carries a limit, removing it lifts the
   limit.
@@ -135,8 +213,103 @@ file carries most of the reasoning behind the tool.
   where Logseq shows them. The body is still one `get-block` away. See
   [#98](https://github.com/muellerei/logseq-cli/issues/98).
 
+### Removed
+
+- `analyze-graph` no longer reports `total_todos`; `tasks` replaces it (#92).
+  A key that changes its value never keeps its old name, so a reader of the
+  old key fails loudly.
+
+- `analyze-journal-patterns` no longer reports `tasks.total_complete`,
+  `tasks.total_incomplete` and `tasks.completion_rate`, and no longer reports
+  `tasks_complete` and `tasks_incomplete` per entry (#92).
+
 ### Fixed
 
+- A change made to a block while a command was changing it was written over
+  without a word. `set-todo-status`, `replace-text`, `set-property` and
+  `remove-property` read the block, build the new text from it and send the
+  whole text back, so an edit made in between (a parallel call of the CLI, a
+  sync, Logseq itself) was lost, and the read-back passed because it compares
+  with what was just sent. Found by an adversarial pass that asked how two calls
+  on one block could interleave; the gap was already on `main`. `update_block`
+  now reads the block again, past the cache, directly before the write when it
+  is told what it replaces, and refuses with the new reason `block_changed`
+  when the text is another, naming the block and writing nothing. An `id::`
+  line is left out of the comparison: a ref written earlier in the same
+  `replace-text` stores the id of a block it then replaces, and that is not a
+  change made elsewhere (measured, #95). The window is smaller, not closed: it
+  is now the requests between that read and the write, the same kind of gap the
+  editor check leaves (#92).
+
+- `get-todos` skipped today when working out `next_due` of an overdue repeating
+  task whose next occurrence falls on today. With today at 2026-10-06, a task
+  last due 2026-09-29 with `+1w` reported 2026-10-13, and one due yesterday
+  with `+1d` reported tomorrow, so `--due-from today --due-to today` missed a
+  task that is due today. Found while checking the fix below against the case
+  one step before it; the error was already on `main`. The single step and the
+  loop now stop at a date that is today or later, as a stored date from today
+  onward already did. Three tests had written the old reading down as "after
+  today" and now say "on or after today" (#92).
+
+- `get-todos` reported `next_due` one repeat too late for a repeating task whose
+  date is today or still ahead. `SCHEDULED: <2026-10-11 Sun +1w>` with today
+  at 2026-10-06 reported 2026-10-18, and `.+1d` dated today reported tomorrow,
+  so `--due-from` and `--due-to` left the task out of the range that holds it.
+  That is the normal state of a task that was ticked off before, because
+  Logseq then writes the next date into the text. Only `++` kept such a date.
+  Found by trying every form with a stored date from today onward; the error
+  was already on `main`. A stored date that is not past is now the next
+  occurrence for `+`, `.+` and `++` alike. A date that is past is unchanged
+  (#92).
+
+- `get-todos` stopped with a traceback and exit 1 on a block whose repeater
+  holds a number no date can reach, such as `SCHEDULED: <2026-01-01 Thu
+  +99999999999d>`, or one of more than 4300 digits. A single such block, typed
+  or pasted, made the command unusable for the whole graph. The task now lists
+  like one whose interval cannot be read: it has no `next_due`, and a
+  `--due-from` or `--due-to` range leaves it out, counts it in
+  `repeating_excluded` and names it on stderr. Found by trying oversized numbers on every unit and form of repeater;
+  the crash was already on `main`.
+
+- `analyze-graph` reported the size of the first cluster it found as
+  `largest_cluster`, not the size of the largest one. The order of the clusters
+  follows the order of the pages, so a graph whose first linked group was small
+  under-reported it (found while reading the code next to the report;
+  reproduced with two groups of 2 and 3 pages) (#92).
+
+- `get-todos` worked out `next_due` of a repeating task from the first repeater
+  in its text, not from the timestamp whose date it reports. A task with
+  `SCHEDULED: <… .+1d>` and `DEADLINE: <… +1w>` reports the deadline, but
+  `next_due` followed the daily repeater, so `--due-from` and `--due-to` could
+  place it up to six days off. Found by working the date out by hand with
+  Logseq's formula. `next_due` now comes from the timestamp whose date is
+  reported, with that timestamp's own repeater; a timestamp without a repeater
+  does not move, so `next_due` is its date (#92). Timestamps are read only from
+  the SCHEDULED and DEADLINE lines of a block, as Logseq reads them; a
+  `SCHEDULED:` in running text or in a code fence is not one.
+
+- A task with a date can be written in one call. `add-note-content --content
+  $'TODO x\nSCHEDULED: <2026-09-25 Fri>'` made two blocks: the `SCHEDULED:` line
+  became a block of its own and had no effect, and the dry run said "2
+  block(s)". Found by reproducing it on a probe page. Logseq's own files keep
+  `SCHEDULED:` and `DEADLINE:` lines and a `:LOGBOOK:` drawer with the block
+  above, as they keep a `key:: value` line, and the outline parser now does
+  too, for every command that reads its text as an outline (`add-note-content`,
+  `add-journal-content`, `add-journal-block`, `insert-block` with an indented
+  outline, `--tree` given as text). A `:LOGBOOK:` that no `:END:` closes is
+  text, not a drawer (measured with Logseq's parser, mldoc 1.5.7), and so is
+  one whose `:END:` comes only after a bullet line, so it hides nothing and
+  swallows no bullet after it (#92).
+
+- `set-todo-status` no longer breaks a heading task: `## TODO ship it` became
+  `DONE ## TODO ship it`, a block Logseq reads no task in, because the command
+  took the first word of the text for the marker. It now swaps the marker where
+  Logseq reads it, behind the heading prefix (`## DONE ship it`), and
+  `get-todos` reports `content` without the marker (`## ship it`). It also keeps
+  the space between the marker and a line break: `TODO \nnotes` became
+  `DONE\nnotes`, which Logseq no longer reads as a task (measured with Logseq's
+  parser, mldoc 1.5.7) (#92).
+
 - A uuid that Logseq holds only as the placeholder of a dead ref (content
   `id:: <uuid>`, no page, kept once Logseq reads the file with the ref
   again) is no block to any command now. `update-block`,