Repository navigation
Releases
1.0.37
1.0.37
中文
修正:关闭计时后一直保持,不再换卡自动恢复
1.0.36 里点眼睛只会关掉当前这张卡的计时,换到下一张卡就自己回来了。实际用起来,干扰往往持续不止一张卡,所以改成:
- 关闭后一直保持关闭,换卡、评完一张又一张都不恢复
- 同一位置留一个小时钟按钮,点它才重新打开;打开后从那一刻重新计时
- 关闭期间不测量、不记分,不会留下错误的 Hard 分
- 重载插件(或重开 Obsidian)后恢复默认显示
请测试
- 点眼睛关闭,连续复习三张卡:胶囊是否一直不出现,位置是否一直留着小时钟按钮
- 点小时钟:胶囊是否在该卡当前位置出现,并从 0 开始跑
- 关闭状态下评一张卡:是否不产生任何超时分
English
Fixed: switching the clock off now lasts
In 1.0.36 the eye button only switched off the current card, and the clock came back on the next one. A distraction usually outlasts one card, so:
- Once off, it stays off across cards and ratings
- A small clock button stays in the same spot; pressing it turns the clock back on, timing from that moment
- While off, nothing is measured and nothing is scored, so no wrong Hard point can be left behind
- Reloading the plugin (or Obsidian) brings the clock back
Please test
- Press the eye, then review three cards in a row: does the pill stay away, with the small clock button still in place
- Press the clock: does the pill appear on that card and count from zero
- Rate a card while the clock is off: does it score nothing
1.0.36
1.0.36
中文
修正:超时的 Good / Easy 不再累计失败连击
原来「评了 Good / Easy 但超时」会让失败连击保持不断、只保留超时那几分。现在改成:
- 超时分只在 Again / Hard 上折算
- Good / Easy 无论用时多少,一律结束这一轮失败连击
- 理由:用对了就是记住了;超时的 Good 多半是中途分心,不该让它留在失败连击里,否则分心反而能"攒"到建子卡
新增:计时胶囊可以关掉
两个胶囊(问题面「作答」、答案面「记忆」)右侧都多了一个眼睛按钮:
- 点它 = 关闭这张卡的计时:不测、也不记分,本次复习不会因为关闭前的时间留下任何 Hard 分
- 关闭后同一位置留一个小时钟按钮,点它重新打开;打开后从那一刻重新计时,关闭期间的时间不计入
- 关闭是「按卡」的:换到下一张卡自动恢复显示,不需要手动再开
请测试
- 把「每步记几个 Hard」设为 1,停到 2 倍预算,评 Good:失败连击是否清零(
+ Child不因此变红) - 同样停到 2 倍,评 Hard:是否记 1 分(
超时:Hard +2) - 问题面点眼睛:胶囊是否消失、出现小时钟按钮;揭示后再点小时钟,是否从 0 重新跑
- 答案面点眼睛再评分:是否不产生任何超时分
- 关闭后进入下一张卡:胶囊是否自动回来
English
Fixed: a slow Good or Easy no longer keeps the failing run
A Good/Easy that ran long used to keep the failing run alive with only its overtime points. Now:
- Overtime points only count on Again / Hard
- Any Good or Easy ends the run, however long it took
- The reason: getting it right is remembering it; a slow Good is usually a distraction, and letting it feed the run would reward wandering off
New: the clock pills can be switched off
Both pills (Answering on the question side, Recalling on the answer side) carry an eye button:
- It turns the clock off for that card: nothing measured, nothing scored, so the time before switching it off cannot leave a Hard point
- A small clock button stays in the same spot to turn it back on; timing restarts from that moment, and the hidden stretch is not counted
- The switch is per card: the next card shows its clock again without any action
Please test
- Set "Hard points per step" to 1, wait until twice the budget, rate Good: does the failing run reset (the "+ Child" button stays calm)
- Wait the same, rate Hard: does it score 1 point (
Too slow: +2 Hard) - Press the eye on the question side: does the pill disappear and a small clock button appear; after revealing, does pressing the clock restart from 0
- Press the eye on the answer side, then rate: does it score nothing
- After switching off, move to the next card: does its clock come back
1.0.35
1.0.35
中文
建议用时按"真的记住了"来算
原来的预算只够"看一眼、想一下",而且只有一条历史权重。这一版把作答和记忆拆开:
作答 = (基础秒数 + 答案字数 ÷ 40 × 每段秒数) × 作答系数
记忆一遍 = (基础秒数 + 答案字数 ÷ 40 × 每段秒数) × 记忆系数
记忆首遍 = (基础秒数 + 首遍基础秒数 + 答案字数 ÷ 40 × 每段秒数) × 记忆系数
记忆 = 记忆一遍 × (记忆遍数 − 1) + 记忆首遍
- 记忆遍数(新设置,默认 3):一共过几遍。首遍已经算一遍,所以后面只再加「遍数 − 1」遍。
- 首遍基础秒数(新设置,默认 20):第一遍是「一遍 + 额外整理思路的时间」,所以首遍 = 基础 + 首遍额外 + 阅读时间。
- 两条权重分开:作答系数默认 −20%,只用于问题面的作答时间(越难给得越少,避免过度纠结);记忆系数默认 +20%,用于记忆时间的每一遍(越难给得越多)。设置里这两行就叫「作答系数(%)」和「记忆系数(%)」。
默认值下的变化(基础 10s、每 40 字 10s、3 遍、首遍 20s):
| 卡片 | 原来 | 现在 |
|---|---|---|
| 20 字短问答 | 15s | 65s |
| 100 字方剂组成 | 35s | 125s |
| 400 字条文 | 110s | 350s |
两点说明:
- 没有上限了。以前的 600 秒封顶会把长答案一律显示成 10:00;现在长答案记三遍就是二十几分钟,照实显示。
- 字数统计忽略
**、==、~~这类强调标记和链接,只算真正要读的字。
作答和记忆是两个计时
- 两个胶囊都在顶部「全部牌组」那一行,位置和只有作答计时时一样。
- 问题面
作答 0:28 / 4:47:从卡片出现到揭示,量的是「一遍」的作答预算,只提醒别在一个词上耗太久。 - 揭示后同一个位置换成
记忆 0:05 / 17:22:从答案出现起重新计时,不接着用作答的时间,量的是记忆正确答案的用时。也带 ↻,点一下从当前时刻重来,评分按重来后的数字算。 - 只有记忆预算超时才记分:作答胶囊超时只是变红,不产生 Hard 分。
- 超时记忆:每多几倍算一步(新设置,默认 0.5):超过预算后,每多半个预算算一步——1.5 倍一步,2 倍两步。
- 超时记忆:每步记几个 Hard(新设置,默认 0.5):每一步给失败连击加这么多 Hard 分。默认下 1.5 倍记 0.5、2 倍记 1、2.5 倍记 1.5;设为 0 就只显示时间、不计分。
- 这些分数写进
fakecard/reviews/的slowPoints字段,和 Hard 的 1 分进同一个计数器;到「失败分阈值」(默认 4)时「建子卡」按钮变红,点它才进入建卡流程,不会弹窗打断。 - 评分 Good/Easy 但超时:失败连击不中断,但只保留超时那几分;正常 Good 仍然清零。
- 没揭示就评分(跳过)不记分;关闭「显示建议作答时间」时整个机制不启用。
请测试
- 问题面:胶囊是否在问题上方、是否写作「作答」、数字是否每半秒跳动
- ↻:问题面点一次是否从当前时刻重计;揭示后再点答案面的 ↻,数字是否重新开始跑
- 揭示后:答案面胶囊是否从 0 开始跑「记忆」,位置和样式与问题面一致
- 设置里分别改「作答时间:评分历史权重」和「记忆时间:评分历史权重」,预算是否一边变少一边变多
- 改「记忆几遍算过关」「首遍基础秒数」,预算是否按公式变化(可对照 tooltip)
- 把「每多几倍算一步」保持 0.5,停到 1.5 倍再揭示:是否显示
超时:Hard +0.5;停到 2 倍:+1 - 把「每步记几个 Hard」改成 1:2 倍时是否变成
+2 - 把它改成 0:答案面仍显示时间,但不再记分
- 连做几次超时后,「+ Child」按钮是否变红
English
The suggested time now covers an actual recall
The old budget only paid for one quick look, and it had a single history weight. Answering and recall are now separate:
Answering = (base + characters / 40 × seconds per chunk) × answering weight
One memory pass = (base + characters / 40 × seconds per chunk) × memory weight
First memory pass = (base + first-pass base + characters / 40 × seconds per chunk) × memory weight
Remembering = one memory pass × (passes − 1) + first memory pass
- Passes (new setting, default 3): how many times the answer is gone over. The first pass already counts as one, so only
passes − 1more are added on top. - First-pass base seconds (new setting, default 20): the first pass is "one pass plus the time to sort the structure out", so it is base + first-pass base + reading time.
- Two weights, separately set: the answering weight defaults to -20% and only touches the question clock (a hard card gets less time, so overthinking is flagged); the memory weight defaults to +20% and touches every memory pass (a hard card needs the time). The Settings rows are "Answering weight (%)" and "Memory weight (%)".
With the defaults (10s base, 10s per 40 characters, 3 passes, 20s first pass):
| Card | Before | Now |
|---|---|---|
| 20-character question | 15s | 65s |
| 100-character answer | 35s | 125s |
| 400-character passage | 110s | 350s |
Two notes:
- There is no upper bound any more. The old 600-second clamp turned every long answer into 10:00; a long answer gone over three times now shows its real twenty-odd minutes.
- Answer length ignores
**,==,~~and links, so it counts only the words to be read.
One pill on both sides
- Both pills sit in the toolbar row with the deck filter, where the old clock was.
- The question side shows
Answering 0:28 / 4:47, measured against ONE pass: the reminder not to linger on a word. - After revealing, the same spot shows
Recalling 0:05 / 17:22: its own clock, started when the answer appears instead of carrying the answering time over, measuring how long the correct answer is being memorised. It has a ↻ too; a rating scores the number after the last reset. - Only memory overtime scores: the answering pill turning red does not earn any Hard points.
- Slow recall: extra budgets per step (new setting, default 0.5): past the budget, every extra half budget is one step - 1.5x is one step, 2x is two.
- Slow recall: Hard points per step (new setting, default 0.5): each step adds that much to the failing run. With the defaults, 1.5x gives 0.5, 2x gives 1 and 2.5x gives 1.5. Set it to 0 to show the time without scoring it.
- Those points are logged as
slowPointsinfakecard/reviews/and counted with Hard's one point. At the failing-score threshold (default 4) the "Child" button turns red; clicking it opens the child-card flow. Nothing pops up on its own. - A Good/Easy that was slow keeps the run alive with only its overtime points; a normal Good still resets it.
- Skipping the card without revealing scores nothing; turning off "Show suggested answer time" disables the whole mechanism.
Please test
- Question side: is the pill above the question, labelled
Answering, with the digits moving every half second - Press ↻ on the question side: does it restart from now; after revealing, press ↻ on the answer side: does the clock restart
- After revealing: does the answer pill start from 0 as
Recalling, in the same place and style as the question side - Change the answering weight and the recall weight separately: does one shrink the budget and the other grow it
- Change passes / first-pass base seconds: does the budget follow the formula shown in the tooltip
- Keep the step at 0.5 and wait until 1.5x before revealing: does
Too slow: +0.5 Hardappear; at 2x,+1 - Set the points per step to 1: is 2x then
+2 - Set it to 0: the time still shows, but nothing is scored
- Do that a few times: does the "+ Child" button turn red
1.0.34
1.0.34
中文
修复:CSS lint 警告 css-display-contents
审核在 styles.css:1287 报:
- Warning:Unexpected browser feature "css-display-contents" is only partially supported by Obsidian 1.11.4
这条规则是 1.0.33 给浏览页批量操作加的那层容器用的——当时为了让「移动牌组 / 暂停 / 删除」既能在自己节点里单独重绘,又能在工具栏里和别的控件排成一行,用了 display: contents。现在改成普通的 display: flex(gap: 8px,与工具栏自身的间距一致,所以视觉上没有任何变化),没有勾选时用 :empty 把它整个收起,不留空隙。
样式表里已无 display: contents。
请测试
- 浏览页勾选几张卡:批量按钮出现的位置与间距是否和之前一样
- 取消全部勾选:工具栏是否正常收拢、没有多出来的空白
- 搜索框、牌组下拉、
New card按钮的排列是否正常
English
Fixed: the css-display-contents lint warning
The review reported this in styles.css:1287:
- Warning: Unexpected browser feature "css-display-contents" is only partially supported by Obsidian 1.11.4
The rule belonged to the container added for the Browse batch actions in 1.0.33: it had to be repainted on its own while still sitting in the toolbar row, and display: contents did both in one declaration. It is now a plain display: flex row with gap: 8px, matching the toolbar's own gap, so nothing moves on screen; when nothing is ticked the node is hidden with :empty instead of leaving an empty gap.
The stylesheet no longer contains display: contents.
Please test
- Tick a few cards in Browse: do the batch buttons appear in the same place with the same spacing
- Untick everything: does the toolbar collapse cleanly, with no leftover gap
- Do the search box, deck dropdown and
New cardbutton still line up
1.0.33
1.0.33
中文
修复:浏览页搜索框按空格之后输不进去
浏览页每敲一个字符都会整块重绘,搜索框的 DOM 节点被删掉重建。旧代码虽然用 focus() 把焦点塞回去,但每换一个节点就会丢光标、丢选区,也会打断中文输入法的组合——空格正是上屏键,所以一到空格就表现为「输入结束」,之后再也敲不进去。
现在工具栏和搜索框只创建一次,按键只重绘下面的结果区(快捷筛选、批量按钮、表格、预览、补全列表):
- 空格、多条件查询(
deck:方剂 tag:麻黄 is:due)都能正常输入 - 光标停在敲字的位置,不会每次按键跳到末尾
- 中文输入法:拼音组合不再被打断,空格上屏正常
- 从
⌄菜单插入条件、点快捷筛选时,光标落到插入内容的末尾
顺带:牌组下拉的选项会在卡片库变化时重建,新建牌组不必重开标签页。
说明:多条件搜索语法
README 中英文两段都补上了完整的筛选语法表,并加了组合示例(deck:x tag:y is:due、tag:x -is:paused、prop:due<7 prop:ivl>=21、added:7 rated:1 等),同时说明空格 = AND、- 取反、带空格的值用 " 包起来。
请测试
- 在搜索框输入
deck:xxx tag:yyy is:due:空格能否连续输入、光标是否停在原处 - 用中文输入法打中文再按空格上屏,能否继续输入
- 点
⌄插入条件、点快捷筛选 chip:搜索框内容与光标是否正确 - 勾选多张卡:批量移动 / 暂停 / 删除按钮是否照常出现,排版有没有变
- 牌组下拉是否列出全部牌组;新建牌组后是否出现
English
Fixed: the browse search box stopped accepting input after a space
Every keystroke repainted the whole Browse view, so the search field was torn down and rebuilt. The old code called focus() on the replacement, but a new node still loses the caret and the selection, and it cuts an input method's composition short - and the space bar is the key that commits a Chinese candidate. Typing therefore ended at the first space.
The toolbar and the field are now built once; a keystroke only repaints what sits below it (quick filters, batch buttons, table, preview, completion list):
- Spaces and multi-term queries such as
deck:方剂 tag:麻黄 is:duetype normally - The caret stays where you typed instead of jumping to the end on every key
- Chinese input methods keep their composition; the space bar commits as usual
- Inserting a term from the
⌄menu or clicking a quick filter puts the caret after the inserted text
Also: the deck dropdown is rebuilt when the card library changes, so a new deck shows up without reopening the tab.
Docs: multi-term search syntax
Both the Chinese and the English part of the README now carry the full filter table plus combination examples (deck:x tag:y is:due, tag:x -is:paused, prop:due<7 prop:ivl>=21, added:7 rated:1), and state that spaces mean AND, - negates and a value with spaces goes in quotes.
Please test
- Type
deck:xxx tag:yyy is:duein the search box: do the spaces go in, does the caret stay put - Type Chinese with an input method and press space to commit, then keep typing
- Insert a term from the
⌄menu, click a quick filter: is the field content and caret right - Tick several cards: do the batch move / pause / delete buttons still appear, is the layout unchanged
- Does the deck dropdown list every deck, including one created just now
1.0.32
1.0.32
中文
撤回:触屏拖动与捏合缩放
1.0.31 给导图加了 pointerdown / pointermove 那一套触屏手势。这一版把它撤掉了,导图回到只认鼠标:
- 按住拖动 = 平移,
Ctrl+ 滚轮 = 缩放,双击 = 回中,和 1.0.30 完全一致 - 触摸板仍然可用:双指滑动产生的是滚轮事件,所以照样能平移和缩放
- 样式表里的
touch-action: none一并删除,页面不再因此禁用触屏滚动
修复:社区审核报的两个问题
审核在 src/mapView.ts 报了两条,来源都是同一处:
- Error:Unsafe assignment to innerHTML(第 247 行)
- Warning:Do not write to DOM directly using innerHTML/outerHTML
原因是节点文本的渲染缓存:为了在遮罩切换时少解析一遍 Markdown,之前把 MarkdownRenderer.render 的结果存进 span.innerHTML,下次同一个词直接写回去。现在整块缓存已经删掉,节点文本每次重绘都交给 MarkdownRenderer.render 重新渲染——不再有任何 innerHTML / outerHTML 赋值。
代价是:遮罩切换、连线重算这类重绘会把可见节点的文本重新渲染一次,超大导图上会略慢。换来的是审核可以过,而且没有把外部 HTML 拼进 DOM 的路径。
请测试
- 鼠标:拖动平移、
Ctrl+ 滚轮缩放、双击回中是否正常 - 触摸板:双指滑动是否还能平移、
Ctrl+ 双指滑动是否还能缩放 - 遮罩切换(点击节点、按住
Alt点选):文字是否仍然正确显示、是否出现闪烁 - 导图节点宽度:短标签是否仍是窄框、长句是否在 420px 处折行
English
Reverted: touch panning and pinching
1.0.31 added a pointerdown / pointermove gesture layer to the mind map. This release takes it out again, so the map is back to being mouse-only:
- Drag to pan,
Ctrl+wheel to zoom, double-click to re-centre - exactly as in 1.0.30 - A trackpad still works: two-finger scrolling arrives as wheel events, so it pans and zooms as before
touch-action: noneis gone from the stylesheet, so the page no longer blocks touch scrolling for the map
Fixed: the two issues from the community review
The review reported two things in src/mapView.ts, both from the same place:
- Error: Unsafe assignment to innerHTML (line 247)
- Warning: Do not write to DOM directly using innerHTML/outerHTML
The cause was the node render cache: to avoid re-parsing the same Markdown on every mask toggle, the output of MarkdownRenderer.render was copied into span.innerHTML and written back for repeated text. That cache is now removed entirely - every repaint hands the node text to MarkdownRenderer.render, and no innerHTML / outerHTML assignment remains anywhere.
The cost: a repaint (mask toggle, edge re-layout) re-renders the visible labels, which is a little slower on a very large map. In exchange the plugin is clean under review and no external HTML is ever spliced into the DOM.
Please test
- Mouse: drag to pan,
Ctrl+wheel to zoom, double-click to re-centre - Trackpad: does two-finger scrolling still pan, and
Ctrl+two-finger still zoom - Mask toggling (clicking a node,
Alt-clicking): do labels still show correctly, is there any flicker - Node widths: short labels still narrow, long sentences still wrap at 420px
1.0.31
1.0.31
中文
新增:触控板 / 触摸屏可以用手指拖动和缩放导图
之前导图只认鼠标:拖动要靠鼠标按键,缩放要 Ctrl + 滚轮。现在:
- 单指(或触摸板单指按住)拖动 = 平移
- 双指捏合 = 缩放(以两指中点为中心)
- 鼠标行为完全不变:按住拖动平移、
Ctrl+ 滚轮缩放、双击回中 - 页面上加了
touch-action: none,手指在图上滑动不会带着整页滚动
内部:两处整理
- 居中逻辑合并:
focusOn与centerOn原本各有一份坐标换算,现在只有centerOn(el, scale)一处,fitScale()单独负责“整图放得下是多大” - 测量方式更稳:节点宽度改用 CSS 的
width: max-content(受max-width: 420px约束),不再需要在测量时把画布临时撑到 8000px。这样节点尺寸不再取决于容器的临时宽度,测量和布局的耦合少了一层
请测试
- 触摸屏或触摸板:单指拖动是否平移、双指捏合是否缩放、松手后是否停住
- 鼠标:拖动、
Ctrl+滚轮、双击回中是否与之前一致 - 导图节点宽度:短标签是否仍是窄框、长句是否在 420px 处折行(不应出现整块变窄或文字竖排)
- 复习与浏览视图里的导图:居中位置是否正常
English
New: pan and zoom a mind map with a touch screen or trackpad
The map used to be mouse-only: drag with a button, zoom with Ctrl+wheel. Now:
- One finger (or a trackpad press-and-drag) pans
- Two fingers pinching zooms around their midpoint
- The mouse behaves exactly as before: drag to pan,
Ctrl+wheel to zoom, double-click to re-centre touch-action: nonekeeps the page from scrolling while a finger drags the map
Internal, two tidy-ups
- Centring merged:
focusOnandcenterOneach had their own coordinate math; now there is onecenterOn(el, scale), withfitScale()answering "how big does the whole map fit" - Steadier measurement: node width now comes from CSS
width: max-content(capped bymax-width: 420px), so the canvas no longer has to be temporarily stretched to 8000px while measuring. A node's size no longer depends on a temporary container width
Please test
- Touch screen or trackpad: does one finger pan, do two fingers pinch, does it stop when you let go
- Mouse: drag,
Ctrl+wheel and double-click still behave as before - Node widths: are short labels still narrow, and do long sentences wrap at 420px (nothing should become a narrow column of characters)
- Mind maps in review and in the browse preview: is the centring still right
1.0.30
1.0.30
中文
内部重构:三个新建面板共用一套“两阶段”逻辑
问答、填空、导图面板之前各自写了一遍“隐藏编辑器 / 显示卡片 / 换按钮文字 / 触发对应回调”,三份代码会慢慢偏离。现在共用 createStages()(src/panelStages.ts),面板之间只剩两点不同:两半各装什么、卡片出现时要做什么。
对使用者的影响:行为不变。上面三个面板的「完成 / 编辑」按钮、切换效果、高度都与 1.0.29 一致。
性能:地图节点文本渲染加了缓存
之前每次重绘都会重建所有节点,并给每段可见文字跑一次 MarkdownRenderer。一张上百节点的图,点一次遮罩就要重新解析整张图。
现在按“源文件 + 文本”缓存渲染出的 HTML(上限 500 条,超出即整体清空)。同一个词再次出现时直接复用,不再解析。对使用者的影响:大图在挖空、改字、拖动后重绘更快;小图感觉不出差别。
请测试
- 三个新建面板:
完成/编辑是否与之前一致(尤其导图面板第一次打开是否居中、编辑后回到导图是否重新居中) - 一张较大的导图(几十个节点):连续点几个节点遮罩、双击改字,是否比之前流畅且位置不乱
- 一张图里同一个词出现在多个节点:是否都能正常显示(缓存复用)
English
Internal: the three new-card panels share one two-stage switch
The Q&A, cloze and map panels each had their own copy of "hide the editor, show the card, swap the button label, run the right callback", and three copies drift. They now share createStages() (src/panelStages.ts); a panel only says what its two halves contain and what must happen when the card appears.
For you: behaviour is unchanged. The "Done / Edit" button, the switch and the sizes are the same as 1.0.29.
Performance: mind map node text is cached
Every repaint rebuilt every node and ran MarkdownRenderer once per visible piece of text, so toggling a mask on a hundred-node map re-parsed the whole map.
The rendered HTML is now cached by source path and text (bounded at 500 entries, cleared as a whole when full). The same word appearing again is reused instead of parsed. For you: large maps repaint faster after a mask, an edit or a pan; small maps feel the same.
Please test
- The three panels: does "Done / Edit" behave as before (the map in particular: centred on first open, and centred again when returning from the outline)
- A large map: toggle a few masks and edit a node - does it feel faster, and does the view stay put
- The same word in several nodes: do they all render (the cache is shared)
1.0.29
1.0.29
中文
只有代码规范改动,功能与 1.0.28 完全一致
社区评审提出的三条:
- 不再直接传
new Component():问答面板与地图渲染都改为先保存到变量,在重绘前和关闭时调用unload(),嵌入笔记的监听器不会累积 - 跨窗口安全判断:
instanceof Element改为 Obsidian 的instanceOf(Element),在弹出窗口里也能正确识别节点 - SVG 用 Obsidian 的构造方法:连线层与连线改用
createSvg(),不再直接调用document.createElementNS
同类写法已全文排查:仓库里不再有 document.createElement、instanceof Element、直接传 new Component() 这三类写法。
请测试
- 打开问答卡的新建面板,在输入与「完成」之间来回切换若干次:渲染是否正常、关闭后是否有异常
- 复习一张地图卡:连线是否照常绘制(这轮改了 SVG 的创建方式)
- 在弹出窗口(把标签页拖出主窗口)里打开复习或浏览视图,地图节点是否仍可点击、遮罩定位是否正常
English
Code-style fixes only; behaviour is identical to 1.0.28
Three community-review items:
- No
new Component()passed inline: the Q&A panel and the map renderer keep it in a field and callunload()before a repaint and on close, so embedded notes do not pile up listeners - Cross-window safe check:
instanceof Elementbecame Obsidian'sinstanceOf(Element), which also holds in a popped-out window - SVG through Obsidian's helper: the link layer and its paths use
createSvg()instead ofdocument.createElementNS
The same patterns were swept across the repository: no document.createElement, no instanceof Element, no inline new Component().
Please test
- Open the Q&A panel, switch between input and "Done" a few times: does it render, and does closing stay clean
- Review a map card: are the connectors still drawn (their creation changed)
- Open review or browse in a popped-out window: are map nodes still clickable and the mask focus still correct
1.0.28
1.0.28
中文
导图卡的计时改成和填空卡完全一致
上一版给导图卡单独写了一套算法(洞里的字 + 8 个字符的上下文),导致同一句话在填空卡上是 11 秒、在导图卡上是 19 秒。
现在两者走同一条路径:只统计这张卡所问的那一组洞里的文字,不含节点标签,也不含整份大纲。整节点遮罩在语法上本来就是一个洞,因此自动按洞的内容计时;一组洞跨多个节点时按总字数计算。
实测:
| 卡片 | 洞内容 | 建议用时 |
|---|---|---|
| 填空 c1 | 客观规律 | 11 秒 |
| 导图 整节点 c1 | 客观规律 | 11 秒 |
| 导图 词级 c1 | 客观规律 | 11 秒 |
| 导图(正文很长、洞很短) | 核心要求 | 11 秒 |
请测试
- 复习一张导图卡与一张内容相同的填空卡:建议用时是否一致
- 一张正文很长、但只挖了两个词的导图卡:用时是否仍在一分钟以内
English
A map card is timed exactly like a cloze card
The previous release gave map cards their own formula (the hidden words plus 8 characters of context), so the same sentence took 11 seconds as a cloze card and 19 as a map card.
Both now go through one path: only the words of the hole group this card asks about are counted - no node labels, no outline. A whole-node mask is a hole in the syntax, so it is measured by its content; a group spread over several nodes adds up.
Measured:
| Card | Hole | Suggested time |
|---|---|---|
| cloze c1 | 客观规律 | 11s |
| map, whole node c1 | 客观规律 | 11s |
| map, word-level c1 | 客观规律 | 11s |
| map with a long label, short hole | 核心要求 | 11s |
Please test
- Review a map card and an equivalent cloze card: are the suggested times the same
- A map card whose body is long but only two words are masked: is the budget still under a minute