| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
| |
First step of #147. I added a cl-defstruct bundling the recall fields the schedulers and the item-data round-trip shuttle around (last-interval, repetitions, ease, failures, meanq, total-repeats), with an ADR comment recording why: the same fields were passed as positional lists in three different orderings, and three call sites re-ordered between them by hand.
The struct is inert here. The item-data round-trip and the schedulers adopt it in the following commits.
|
| |
|
|
|
|
| |
I renamed several cryptic locals to say what they hold: idx to index-var in the pop-random gensym, val to raw-value in the five DRILL_* property getters, q to quality in hypothetical-next-review-dates, m to marker in free-markers, the a/b sort and filter lambdas in order-overdue-entries to entry/other, and dat to card-def in the empty-card-type check.
These are pure renames with no behavior change. Byte-compile and the full unit suite stay green.
|
| |
|
|
| |
sym1/sym2 gave no hint of their role. current-meter-char toggles per meter wrap and alternate-meter-char is its inverse, which is what the two make-string calls actually consume.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
A batch of self-contained user-facing improvements, squashed from the feat/org-drill-solo-features branch.
I added an undo for the last rating (issue #2 follow-up). The rating prompt now takes an undo key (org-drill--undo-key, default u): it restores the previous card's scheduling snapshot, drops the recorded quality, and re-queues that card, then returns to the current prompt. Each rating snapshots the scheduling properties and SCHEDULED line onto a per-session stack capped at org-drill-undo-limit (default 3). org-drill-reschedule loops on the rating read so undo doesn't rate the current card.
I made the five session-control keys (quit, edit, help, skip, tags) defcustoms so they can be rebound from customize-group (issue #35), keeping their defaults. The 0-5 rating keys stay as-is, since they're tied to the quality scale rather than being variables.
I lifted the hardcoded 100-line entry-text limit in org-drill-get-entry-text into the org-drill-entry-text-max-lines defcustom, defaulting to 100.
I also deleted a commented-out old org-entry-empty-p that the real definition had already replaced.
Existing tests stay green and each change added its own, including snapshot/restore and prompt-loop tests for undo.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
A cleanup pass over org-drill internals, squashed from the refactor/wave3-cleanup branch. No behavior change. Each step kept the existing tests green and added its own.
I shared two duplicated helpers across the language card getters: org-drill--read-property-string and org-drill--face-from-alist.
I factored the cloze body-scan out of the two multicloze presenters into org-drill--cloze-body-bounds, org-drill--count-cloze-matches, and org-drill--hide-cloze-by-index, so each presenter just picks which indices to hide.
I pulled the presenter resolution and the four-way result classification out of org-drill-entry-f into org-drill--resolve-presenter and org-drill--classify-presentation-result, untangling the pivot of every drill iteration.
I split the 37 defcustoms (and the three cloze faces) into four customize sub-groups (display, algorithm, session, leech) so customize-group org-drill is navigable. There's no leitner group because the Leitner settings are defvars.
I documented the 22 defuns that had no docstring, rewrote the corrupted org-drill-presentation-prompt-in-mini-buffer docstring, and switched eleven docstrings to the imperative "Return" (issue #2).
|
| |
|
|
|
|
|
|
| |
org-drill-add-cloze-fontification ran on org-font-lock-set-keywords-hook, which fires in every org buffer, and pushed the cloze rule into org's global org-font-lock-extra-keywords. The cloze regexp is built from the [ and ] delimiters, so an org priority cookie like [#A] matched the cloze pattern and got fontified as a cloze in every org buffer, colliding with org's headline fontification and stripping the heading's org-level-N face.
I replaced the global install with org-drill-mode, a buffer-local minor mode that adds the cloze keywords only to its own buffer via font-lock-add-keywords. org-drill-auto-enable-mode (default on) turns the mode on from org-mode-hook in buffers that hold drill cards, so existing drill files keep their cloze highlighting while plain org buffers stay clean. Highlighting still respects org-drill-use-visible-cloze-face-p.
The cloze regexp itself is unchanged, so the single-line cloze constraint from #38 is preserved.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A batch of test-coverage and hardening work, squashed from the test-work branch.
Tests: deduplicated a colliding leitner-capture test name so make test-name loads again. Added SM2 assert-failure cases, the six basic multicloze variant delegations, the three English-side spanish-verb branches, and org-drill-current-scope branch coverage.
Fix: the entry-point commands (org-drill itself, cram-tree, tree, directory, resume, relearn-item, strip-all-data, merge-buffers) carried no autoload cookies, so M-x failed from a fresh install until something pulled the file in. They're autoloaded now.
Perf: org-drill-shuffle was quadratic because it indexed a list with elt on every swap. It runs a linear Fisher-Yates pass over a vector now, and it checks its argument is a list.
Feat: added org-drill-version, a constant plus an interactive command, so a bug reporter doesn't have to open the file header.
Refactor: extracted org-drill--format-tense-mood, shared by the two verb-conjugation presenters that each carried a copy.
Docs: explained the SM8 magic numbers in the simple8 helpers as empirical fits rather than tunable knobs.
|
| |
|
|
|
|
|
|
|
|
| |
Cask's upstream has slowed. Eask is the actively maintained successor.
Eask's `package-file` directive doesn't auto-install the deps from the .el header's Package-Requires, so the Eask file mirrors emacs/seq/org/persist explicitly. `eask install-deps` also doesn't pull transitive deps, so dash, m-buffer, and shut-up needed their own `depends-on` lines for undercover and elisp-lint to activate.
The Makefile swaps are mechanical: $(CASK) → $(EASK), `cask install` → `eask install-deps --dev`, `cask build` → `eask compile`.
The URL in org-drill.el's header pointed at the abandoned upstream's GitLab issues page. Eask cross-validates that against website-url, so I updated it to the GitHub mirror — where users file issues now.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Two extractions out of the 58-line cond:
- org-drill--entry-empty-and-not-empty-friendly-p: encapsulates the
ugly nested predicate that means 'body is empty AND the card
type doesn't opt in to empty bodies'. Used to be 5 inlined
lines including a let* and an alist lookup.
- org-drill--classify-status: takes the precomputed DUE and
LAST-INT and walks the cond. The decision tree is now a
flat sequence of one-line clauses.
org-drill-entry-status itself drops to 8 lines and reads as
'compute the inputs, classify, return triple'. The :failed branch
also uses org-drill--quality-failed-p instead of inlining the
threshold check.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
merge-buffers was 89 lines of dense cross-buffer marker work mixing
hash-table population, source-buffer iteration with embedded
property reads, scheduling-data writes, and cleanup of unmatched
DEST entries.
Extracted four helpers:
- org-drill--build-dest-id-table: scan DEST and populate the id→
marker table.
- org-drill--copy-scheduling-to-marker: read the current entry's
scheduling state and write it at MARKER (skipping never-rated
items via total-repeats=0 guard).
- org-drill--migrate-from-source: walk SRC, dispatch to the copy
helper or org-drill-copy-entry-to-other-buffer for new items.
- org-drill--strip-unmatched-dest-entries: clean up DEST entries
that have no SRC match.
merge-buffers itself drops from 89 lines to 19 and reads as a
sequence of named phases.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
org-drill was 137 lines mixing five distinct concerns: an org-version
warning (now dead under the org>=9.6 floor), session reset, entry
collection, drill execution, and post-session messaging.
Extracted four helpers, each with a single responsibility:
- org-drill--prepare-fresh-session: zero out queues + counters
- org-drill--collect-entries: scan + sort overdue
- org-drill--queues-empty-p: predicate for the no-pending branch
- org-drill--show-end-message: dispatch resume-hint vs final-report
Plus org-drill--show-resume-hint for the keystr-aware suspended
message. Removed the dead org<7.9.3f warning block (the org>=9.6
floor makes it unreachable).
org-drill itself drops from 137 lines to 36 and the cl-block
wrapper goes away — the cl-return-from inside org-drill-entries
returns through the normal control flow now.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
org-drill-final-report was 95 lines, dominated by two big format
calls (the main summary and the low-pass-rate WARNING) plus an
inlined queue-tag pattern that propertized 5 different counts in
the same shape.
Extracted three helpers:
- org-drill--queue-tag: builds a propertized 'N label' string for
one queue (failed / overdue / new / young / old). Replaces 5
inlined propertize calls.
- org-drill--build-final-report-summary: takes the session, and
returns the formatted main summary string.
- org-drill--build-low-pass-warning: takes the session and pass-
percent, returns the formatted warning string.
org-drill-final-report becomes a 12-line orchestrator that wires
the helpers together with the wait-and-read loop.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The main loop body reached 7 levels of indentation in the success
path: while > destructuring-bind > save-excursion > cond > t-clause
> let > cond > inner-cond.
Extracted two helpers:
- org-drill--pick-next-marker: chooses between resuming the
current-item slot and popping a fresh marker, returning a (marker
. next-resuming-p) cons so the caller updates resuming-p in
lock-step.
- org-drill--route-rating-result: routes the rating result into
the session's again/done queues and returns a symbol (quit/edit/
skip/next) telling the caller whether to break the loop.
org-drill-entries shrinks from 54 lines to 19, and the deepest
nesting drops from 7 levels to 4.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
leitner-rebox
The two interactive rating loops (reschedule and leitner-rebox)
shared roughly 60 lines of identical code: the same key-prompt
string, the same memq-based exit-on-rating loop, the same arrow/
scroll/wheel cond, the same tags-key dispatch, and the same help-key
toggle. Only two things differ — the rating-explanation help text
and the per-rating action that follows — and the comment in
leitner-rebox literally said "All this is shared with drill-
reschedule. And what does it do?"
Extracted org-drill--read-rating-key with two args: typed-answer
(for the typed-answer flow's 'Your answer: ...' line) and
rating-help-block (the multi-line ratings explanation specific to
the scheduler).
Both call sites collapse from ~50 lines of inlined loop to a single
call. reschedule and leitner-rebox now consist of just the rating-
specific actions plus this read-key call.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Five presenters opened with the same three-deep wrap:
(org-drill-with-hidden-comments
(org-drill-with-hidden-cloze-hints
(org-drill-with-hidden-cloze-text
...body...)))
Combined into org-drill-with-card-display. Five sites
(present-simple-card, present-simple-card-with-typed-answer,
present-two-sided-card, present-multi-sided-card,
present-spanish-verb) lose 2-3 lines of nesting each.
Multicloze-hide-n / hide-nth use a different envelope (only two of
the three wraps; they hide specific clozes by index, not all of
them) so they keep their explicit nesting. Same for
present-card-using-text, which substitutes with-replaced-entry-text
for with-hidden-cloze-text.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The check (<= quality org-drill-failure-quality) appeared in five
places: SM2/SM5/Simple8 schedulers, smart-reschedule, and the
main org-drill-entries loop body. Each call site does different
things on failure (reset interval, push to again-entries, etc.) so
only the predicate was duplicated.
Extracted as org-drill--quality-failed-p with a docstring naming
the threshold and the role. Five inlined comparisons collapse to
five named predicate calls.
|
| |
|
|
|
|
|
|
|
|
|
| |
The seven-line propertize+concat block that prepends a red leech
warning to the prompt was inlined in three prompt builders:
presentation-prompt-in-mini-buffer, presentation-prompt-in-buffer,
and presentation-prompt-for-string.
Extracted org-drill--maybe-prepend-leech-warning as a single helper
the three call. 21 lines duplicated → one definition + three
one-line call sites.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Four dead-code blocks deleted, all of them commented-out
alternative implementations next to the live versions:
- 5-line commented (and (>= quality 4) ...) cond branch in the
SM5 scheduler
- 6-line commented (loop do (re-search-forward ...)) alternative
in present-multicloze-hide-n
- 4-line commented unless-error guards at the top of entry-f
- 3-line commented (:tomorrow ...) case branch in
map-entry-function
Prose commentary kept (license, package overview, recent-fix
explanations). Pure deletion, no behavior change.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
org-drill-present-spanish-verb had a 6-branch cl-case where each
branch differed only in two values: which subheading to reveal
(Infinitive or English) and which prompt string to show. Inlined
into a 50-line block of nearly-identical setq calls.
Extracted the (reveal . prompt) pairs into a defconst alist and
reduced the dispatcher to a random-pick + apply-pair pattern.
Function drops from 51 lines to 14. Existing branch tests still
pass — they mock cl-random to a specific index, and the alist's
order matches the previous case order.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The quality-percentage formula
(round (* 100 (cl-count Q qualities)) (max 1 (length qualities)))
appeared six times in org-drill-final-report (once per recall
quality 0..5), each time inlined verbatim. Extracted to a single
helper.
Six call sites collapse from three lines each to one. Behavior
identical (the helper documents the (max 1 ...) divisor as an
empty-list guard). Final-report drops from 95 lines to ~80.
|
| |
|
|
|
|
|
|
|
|
|
| |
The function had an Org-version branch with a legacy <9.6 path that
used (substring (cdr org-time-stamp-formats) 1 -1). Modern Org's
format-strings dropped the angle brackets that the substring slice
assumed, so the legacy path was both dead-code (unreachable under
the org>=9.6 dep floor we just declared) and silently buggy if it
ever did run.
Function is now a one-liner around the modern primitive.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Two dead branches removed:
- (when (version< org-version "9.2") ...) advice on org-get-tags,
plus the org-drill-get-tags-advice helper that backed it. Org 9.2
shipped the new arity in 2018 — well below our org>=9.6 floor.
- (when (= 8 (car ...)) ...) Org 8.x defalias shim that wrapped
org-latex-preview around org-preview-latex-fragment. Org 8 hasn't
been a target for years and is below the org>=9.6 floor.
Both are now genuinely dead after the dep bump in 75b1601.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Three large dead-code blocks removed:
- 16-line commented-out org-drill-entry-due-p (replaced by current
implementation that takes a session arg)
- 17-line commented-out org-drill-hide-all-subheadings-except (the
body was a placeholder docstring + commented body; the real
function lives elsewhere now)
- 25-line block of commented-out alternative
org-drill-add-cloze-fontification + add-hook XXX commentary
Pure deletion, no behavior change. Version control has the history.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
org-drill calls org-fold-show-entry and org-fold-show-subtree from
seven sites without fboundp guards. Both functions arrived in Org
9.6. But the package declared org 9.3 (Package-Requires) /
org 9.2 (Cask), so users on older Org would silently void-function
at runtime instead of getting a clear install-time mismatch error.
Bumped both declarations to org 9.6. Wrapping each of the seven
call sites with fboundp would be the alternative, but Org 9.6 was
released October 2022 — three-and-a-half years ago — and we already
have a follow-up TODO to drop the legacy time-to-inactive fallback
that this version bump unblocks.
Two tests verify the declared dep and that the org-fold APIs are
actually bound on the running Org version.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
User reported that running org-drill on a buffer with a new (no-ID)
entry threw 'Wrong Type Argument: hash-table-p, nil' and stopped the
scan — every subsequent entry was silently skipped, so the user had
to re-run org-drill once per item (10 items meant 10 invocations).
The exact source of the hash-table error is environment-dependent
(Emacs version, Org version, lazy org-id-locations init, Doom
overrides), so this fix targets the user-visible failure mode
instead of the underlying triggering condition.
Wrapped the per-entry body of org-drill-map-entry-function in
condition-case. An error on one entry now logs a 'skipping' message
and the scan continues to the next entry. The session collects all
the well-formed items, and the user can re-run drill once total to
process them — no more once-per-item.
Two regression tests: one verifies the resilience behavior directly
(fail entry 1, scan continues to entry 2), the other documents the
ID-creation-with-uninitialized-locations scenario as a smoke check.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The inner match was [[:cntrl:][:graph:][:space:]]+?, which silently
includes newline. A stray [ could match all the way to a ]
several lines later, covering org headings in between with the
visible-cloze face. Reporter saw lines 4 and 5 of test.org lose
their org-level-N face and use default instead.
Switched the inner class to [^\n]+?. Clozes now stay within a
single line, which matches the design intent and stops the face
bleed. Three new tests cover the regression.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Issue #44 (2021): running org-drill in a TTY emacsclient (the
reporter mentioned tmux) raised "Window system frame should be
used" because LaTeX preview helpers (org-latex-preview,
org--latex-preview-region) require a window system and weren't
guarded.
Wrapped both call sites with (when (display-graphic-p) ...).
- org-drill--show-latex-fragments: now a silent no-op on TTY
- present-default-answer's clear-and-preview block: same guard
LaTeX previews are inherently graphical. The right behavior on
TTY is to skip the preview rather than crash the session — TTY
users still see the underlying source text just fine.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When a user interrupted a drill session to edit or capture, the
session's end-pos slot got set to a marker (or :quit). The end-of-
org-drill cond branched on end-pos: if set, show resume message and
skip org-drill-final-report.
That worked for the first interruption. But on org-drill-resume,
the session was reused with end-pos still carrying the prior marker.
Even when the resumed session completed normally, the same cond
branch fired again — silently skipping final-report.
Clear end-pos at the top of org-drill when resume-p is non-nil, per
Markus's proposed patch on the upstream issue. The resumed session
can now reach the final-report branch.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
org-drill--setup-display saved buffer-local state (mode-line,
variable-pitch-mode) into global defvars and called setq-local on
the current buffer. org-drill--restore-display read those globals
and ran setq-local against whatever buffer happened to be current
at restore time.
If the user switched buffers mid-session, the restore wrote to the
wrong buffer — leaving the original drill buffer's mode-line still
hidden and trampling the destination buffer's mode-line with
whatever was saved from elsewhere.
Captured the buffer at setup in org-drill--saved-display-buffer.
Restore now wraps mode-line and variable-pitch restoration in
with-current-buffer against that saved buffer. Text-scale stays
global (the underlying face attribute is process-wide).
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Issue #45 (2021): persist-load raised End of file during parsing
at persist.el:413 in some configurations, likely from a corrupted
persist data file. Pre-fix, this propagated up through the
top-level (persist-defvar org-drill-sm5-optimal-factor-matrix ...)
form at file-load time and broke the entire package's load.
Wrapped the persist-defvar form in condition-case. On failure, the
matrix falls back to a fresh nil binding via plain defvar, and a
message tells the user what happened. org-drill continues to load
normally.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Both functions bound session to org-drill-last-session and immediately
called setf / org-drill-entries-pending-p on it without checking for
nil. First-time invocation (or after Emacs restart with no active
session) threw an obscure eieio-oset / nil-slot type error instead
of a clear message.
Added (unless session (user-error ...)) at the top of each function.
A user running M-x org-drill-resume cold now sees a sensible message
telling them to run org-drill first.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
drawer-end was captured as
(save-excursion (re-search-forward ':END:' end t) (point))
which always returns a number — (point) is always defined. The
subsequent (when drawer-end ...) guard was dead, so a malformed
drawer (typo in :END:, mid-edit truncation) ended up with a junk
overlay covering whatever range point happened to land in.
Captured the search result itself and gate on it. Malformed
drawers are now skipped silently; well-formed drawers still get
their normal overlay.
|
| |
|
|
|
|
|
|
|
|
|
| |
When the property is absent, org-entry-get returns nil and
string-to-number errors with wrong-type-argument. Reachable when a
user removes the property mid-session, or when a Leitner-tagged
entry is rebox'd before its DRILL_LEITNER_BOX has been set.
Wrapped the org-entry-get with (or ... "0"). Box 0 makes the rating
semantics still sensible: a downgrade stays at 0, a promotion goes
to 1.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The card-type alist mapped translate_number to a function that no
longer exists in the file. Cards with DRILL_CARD_TYPE: translate_number
crashed with void-function during drill instead of being skipped.
Reporter (issue #43, 2021) said they had old decks using the
documented translate_number type and were getting the crash on
restore. The function was apparently removed at some point without
clearing the alist entry.
Removed the alist entry so entry-f's no-presentation-fn branch fires
and returns skip after messaging the user. Legacy decks now degrade
gracefully instead of crashing the session.
Tests in tests/test-org-drill-translate-number-regression.el lock the
behavior in (entry-f returns skip on translate_number, alist no
longer carries the entry).
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The warning branch divided 100*overdue by (dormant+due) without
guarding the denominator. When both counts are zero — degenerate
scopes (cram with no items collected, pure-failure session on empty
queues) — the call hit arith-error before the warning even rendered.
Wrapped the divisor with (max 1 ...). In the zero case the
percentage reads as 0% rather than crashing the session wrap-up.
Resolves a long-standing pre-existing TODO entry.
|
| |
|
|
|
|
|
| |
Every drilled card was logging "[debug] org-drill: at marker position N"
to *Messages* and flashing it in the minibuffer. Pure noise — the
print statement was clearly a leftover from diagnostic work that
never got cleaned up. Delete it.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The function takes `days-ahead' as &optional, but the schedule cond
called `(= 0 days-ahead)' and `(cl-minusp days-ahead)' before any
type-guard, so passing nil crashed with a wrong-type-argument error.
Today's two callers (the rating-confirmation flow and the
org-drill-relearn-item helper) always pass a number, so this was
latent — but a third caller relying on the documented &optional
shape would hit it immediately.
Switched the cond to require numberp before the value comparisons,
and the default branch now falls back to the algorithm-computed
next-interval when days-ahead is nil. That matches the intent
implied by the optional signature and the docstring.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Plus a docs fix to org-drill-order-overdue-entries' header comment.
16 ERT tests covering:
- org-drill-entry-status: non-drill nil, empty entry nil, virgin :new,
future :future, low-quality :failed, due+short-interval :young,
due+long-interval :old, very-overdue :overdue, skipped-leech
:unscheduled, three-element return shape
- org-drill-entry-days-since-creation: with DATE_ADDED, missing without
flag (nil), missing with use-last-interval-p flag (overdue+interval)
- org-drill-order-overdue-entries: empty stays empty, non-lapsed
sorted by DUE desc, lapsed split (by DUE crossing threshold, not AGE)
appearing after sorted by AGE desc
Fixed misleading header comment at line 2888 — it claimed the lapse
split was by AGE, but the code uses DUE (cl-second). This matches
the semantic gate in org-drill--entry-lapsed-p, so the code was
right and the comment was stale. Updated the comment to state the
actual three-step sort.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
org-drill-hide-cloze-hints checked (null (match-beginning 2)) to detect
"no hint present," but the cloze regex's hint group is an empty-allowed
alternation — the group always participates in the match, so
match-beginning is always a position, never nil.
For a card like "[Paris]" (no hint), the function fell through to
org-drill-hide-region with start = end and made a zero-width overlay.
Cosmetically harmless but accumulates one stray overlay per hint-less
cloze. On a buffer with many such cards the tracking cost is real.
Switched the guard to (= (match-beginning 2) (match-end 2)) — empty match.
Found while writing tests; locked in by tests/test-org-drill-hide-show.el's
test-org-drill-hide-cloze-hints-no-hint-no-overlay.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
kqr (2019-07-22) reported that drill entries whose answer lives inside a child sub-heading were silently skipped. Their example: a question in the heading text and the answer under `** The Answer`. The function returned t (empty) for such entries, so they never got presented during drill sessions.
The cause is `(outline-next-heading)` in `org-drill-entry-empty-p`. That primitive lands on the first heading at any level, including children. So the search range was metadata-end up to the child's heading line, which excluded the child's body. Bodies that lived in child sub-headings never got searched.
I switched the bound to `(org-end-of-subtree t t)`, which covers the whole subtree of the current heading and degrades gracefully at the last heading in the buffer. The reporter suggested `outline-forward-same-level`, but that primitive errors at the last sibling, which would be its own regression. `org-end-of-subtree` is the canonical Emacs idiom for this kind of bound and handles end-of-buffer correctly.
I added `tests/test-org-drill-entry-empty-p.el` with 6 ERT tests across Normal, Boundary (kqr's exact fixture), and edge categories. The two regression tests fail at HEAD before the fix and pass after.
One semantic note worth flagging: any subtree content now counts as non-empty, including bare child headings with no body of their own. The bug report is silent on that case and I expect it to be rare in practice. If anyone reports the new behavior as a regression, the fix would be to filter heading lines out of the graphical-character search.
|
| |
|
|
|
|
|
|
|
|
| |
Two reports from breadncup (issue #52 in 2023, issue #58 in 2024) said that running an org-drill session silently nulled out their `default-input-method`. The reproduction is exact: every rating prompt clears the user's persistent setting.
The cause is `(set-input-method nil)` in `org-drill--read-key-sequence`. When `current-input-method` is nil, calling `set-input-method` with nil clears `default-input-method` as a documented side effect. The unwind-protect on the way back has the symmetric problem, since it passes the captured nil. The fix is to use the primitives that are scoped to current state. `deactivate-input-method` and `activate-input-method` don't touch `default-input-method`, and I wrap each call in a guard so the function is a no-op when no input method is active.
The same pattern lives in `org-drill-response-get-buffer-create`, which propagates the caller's input method into the response buffer. When the caller has no input method active, the captured value is nil and `(set-input-method nil)` runs in the new buffer, clearing `default-input-method` again. I applied the same guard there.
I added `tests/test-org-drill-read-key-sequence.el` with 6 ERT tests across Normal, Boundary (the bug case), and Error categories. The four regression tests fail at HEAD before the fix and pass after.
|
| |
|
|
|
|
|
|
| |
The Simple8 failure branch was missing (cl-incf totaln) while SM2 and SM5 both increment total-repeats on failure. After this change, DRILL_TOTAL_REPEATS counts every review attempt regardless of which scheduling algorithm produced it, including failures.
Going-forward only. Historical totaln values for Simple8 failures stay under-counted by one. Correct counting starts with the next failed review.
Paired with the test commit 5c68f1e, which captured the new expected behavior first. Full suite at 214 of 214.
|
| |
|
|
| |
(cherry picked from commit eacb6d0c018839d8207ee80e02b46b314278ac3f)
|
| |
|
|
| |
(cherry picked from commit 76d45fb0ea6e216b2cb173bdcf73ef284d350ff8)
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
org-drill-entry-days-since-creation
Wrapped org-time-stamp-to-now call in condition-case to gracefully handle
malformed DATE_ADDED property values. Now returns nil instead of crashing
when encountering invalid timestamp formats.
Changes:
- Added condition-case around org-time-stamp-to-now (lines 2896-2898)
- Returns nil on error, allowing the function to fall through to other
branches or return nil gracefully
This prevents unhandled errors in long-running sessions when drill entries
have corrupted or manually-edited timestamp values.
|
| |
|
|
|
|
|
|
|
|
| |
Changed incorrect property names to match standard naming convention:
- LAST_QUALITY → DRILL_LAST_QUALITY (lines 3394-3395)
- LAST_REVIEWED → DRILL_LAST_REVIEWED (lines 3397-3398)
This ensures consistency with the rest of the codebase where all drill
properties use the DRILL_ prefix. The old names would create properties
that don't match the standard schema and wouldn't be read correctly.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Created org-drill-lapse-threshold-days defcustom (default 90) to replace
hardcoded values scattered throughout the code. This improves maintainability
and allows users to customize when entries are considered lapsed.
Changes:
- Added defcustom org-drill-lapse-threshold-days (line 660-669)
- Updated org-drill-order-overdue-entries to use variable (line 2867)
- Simplified org-drill--entry-lapsed-p to use variable (line 2884-2886)
- Added safe-local-variable declaration (line 687)
- Updated docstring references to use variable name
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Replaced switch-to-buffer with with-current-buffer to avoid changing
visible buffers during merge operation. This prevents window state
corruption and allows the function to work correctly in batch mode.
Changed line 3374-3375 from:
(save-excursion
(switch-to-buffer (marker-buffer marker))
...)
To:
(with-current-buffer (marker-buffer marker)
(save-excursion
...))
|
| |
|
|
|
|
|
|
| |
Removed commented-out code that redundantly set the failures variable.
The failures count is already obtained from org-drill-entry-failure-count
on line 1538, making the commented code (lines 1548-1550) unnecessary.
This cleans up maintainability issues and removes confusing dead code.
|
| |
|
|
|
|
|
|
|
|
|
| |
Previously used (or (cdr (org-get-property-block)) (point)) which could
return invalid position if no property block exists. Now properly positions
after heading and metadata using org-end-of-meta-data when property block
is missing.
Affects:
- org-drill-present-multicloze-hide-n (line 2267)
- org-drill-present-multicloze-hide-nth (line 2345)
|