fail to input a sentence nor a Chinese character

[apua]

:editor: Vim 8.0 :os: Mac

I failed to input a sentence nor a Chinese character.

Failed to input a sentence

The state of typing "ji3g4su3184fm4a94cji3t96" is as shown below

It doesn't translate to Chinese characters in addition to first term.

Failed to input a Chinese character

After input a term of phonetic symbols as below, I want to select one to output

However, neither input "space" nor "enter" are expected result:

• input "enter" -- get newline

• input "space" -- doesn't translate

---

If the fact is I misunderstand the usage of ime-phonetic.vim , then the problem might be improvement of guide.

[pi314]

由於這個 repo 是中文輸入法，目前也還沒有英語人士加入，我先用中文回應，確保意思不會被誤解~

---

@apua: 我在 ime.vim 裡面做了以下的 mapping:

inoremap <expr> <CR> (pumvisible() ? "<C-Y>" : "") . "<CR>"

這會使得 <CR> 無論如何都會輸出一個換行

如果改成

inoremap <expr> <CR> (pumvisible() ? "<C-Y>" : "<CR>")

的話，就會符合你的預期，也比較符合 vim 的預設行為

當初會特別去 mapping 它是要修某個 bug，但已經忘記確切的原因了，我去 ime.vim 那裡調整一下

---

vim 的 popup menu 對 [<Up> / <Down>] 和 [<C-n> / <C-p>] 的反應不太一樣：

• 按下 <Up> / <Down> 以後，選項並不會真的插入編輯區，要按下 <CR> 才會
• 按下 <C-n> / <C-p> 以後，選項立刻就會插入編輯區

你覺得把 <Up> / <Down> remapping 成 <C-n> / <C-p> 是個好主意嗎？

[apua]

:editor: Vim 8.0 :os: FreeBSD

• select character with <C-n>/<C-p> -- works fine
• select character with <Up>/<Down> and press <CR> (enter) -- doesn`t add newline , unlike behavior on Mac

[apua]

I think it is unnecessary to remap <Up>/<Down> currently.

Writing down the usage is more important; basically, software cannot be tested without specification.

Besides, in my opinion, making inputting sentences quickly with less typing has higher priority than remapping <Up>/<Down>.

For inputting a single Chinese character, I think the issue could be closed; for inputting sentences, what do you think?

-- ps: why I type in English is not for another people but just a practice

[pi314]

Before your second comment (the :os: FreeBSD one), I changed the <CR> behavior in ime.vim, so please check if ime.vim version is 0.30.2 on both side~

I updated <Up> / <Down> / <CR> and <C-n> / <C-p> information in README, and decide not to remapping <Up> and <Down> to let user have choices.

---

Currently I have no god idea for inputting sentences. Even inputting words (詞) is implemented by add a huge database now. Learning (more frequent words float to the top) is another topic I have no idea now 😛

[pi314]

Maybe this helps: https://blog.timc.idv.tw/posts/javascript-chinese-ime/

Sentences-inputting is a must-have feature. To implement this, maybe we need to add word occurrence frequency information into database.

[pi314]

It comes to me that rime is server/client architecture, perhaps it's the right way to go if we want to do smart-input (long sentences, learning, etc.)

[pi314]

jieba has word occurrence frequency information: https://github.com/ldkrsi/jieba-zh_TW/tree/master/jieba /dict.txt

[pi314]

Planning to rebuild database.

Data source:

• jieba-zh_TW - MIT (https://github.com/ldkrsi/jieba-zh_TW/tree/master/jieba/dict.txt)
• moedict - CC0 (https://github.com/pi314/phonetic-table)
• (maybe) McBopomofo - MIT (https://github.com/openvanilla/McBopomofo/tree/master/Source/Data)
  • https://github.com/openvanilla/McBopomofo/blob/master/Source/Data/BPMFBase.txt
  • https://github.com/openvanilla/McBopomofo/blob/master/Source/Data/BPMFMappings.txt
  • https://raw.githubusercontent.com/openvanilla/McBopomofo/master/Source/Data/phrase.occ

Extract pronouncing information from moedict and McBopomofo. Extract word frequency information from jieba and McBopomofo.

[pi314]

[Status] Rebuilding database with McBopomofo data: https://github.com/pi314/phonetic-table/commit/13c1e0b2435cd9780479cd31d8a87d1e5affb389 This time I write a Makefile to make sure that it is rebuildable and patchable in the future. Not finished yet, but it's closed.

[pi314]

[Planning] After database built, a new word selection algorithm should be implemented.

---

The current algorithm is greedy. For example, given phonetic codes:

A B C D E

It first tries A; if ok, try AB; if ok, try ABC, and so on. If in anyone step the word does not exist, e.g. ABC does not exist, then previous result is used (i.e. AB). The remaining phonetic codes (C D E) is used to search all over again.

One of the cons of this method is that if AB, ABC and CDE are valid words, CDE will never be considered because ABC always being picked up.

---

With word occurrence information, a smarter algorithm is possible. Instead of searching all possibilities (O(2^n)), I think we can use DP (Dynamic Programming). The problem will be reduced to the rod-cutting problem (O(n^3)).

[pi314]

YAY f6122e0

[pi314]

Closed by #17. Finally, this plugin begins to be usable.

@apua I think you can input sentences now. The sentences will not be accurate, but it's another issue. At least ㄨㄛˇㄕˋㄋㄧˇㄅㄚˋㄑㄩˋㄇㄞˋㄏㄨㄛˇㄔㄞˊ really generates 我是你爸去賣火柴 with current database :tada: