Phrase construction or enable_encoder/table_translator documentation?

[erikli]

I want to know how full codes for phrases get converted to shorter codes and couldn't figure it out after trying to read https://github.com/rime/home/wiki/RimeWithSchemata. Is this explained somewhere?

for example in cangjie5:
身體健康 is originally 竹難竹、月月廿田廿、人弓大手、戈中水 and is shortened to become 竹田人手水 and I really don't understand why that happens.

What are the rules? Thank you for any help!

[erikli]

雖然用英文回答會讓我多容易明白，用中文回答也很幫助。非常感謝

[LEOYoon-Tsaw]

In cangjie5.dict.yaml, there's a section of encoder/rules. Those word encoding rules are defined there.
Capital letters locate the character in the word, small letters locate the code to use. While A, B, C count from 1, 2, 3, Z, Y count as -1, -2.

[erikli]

thank you!

[lotem]

It's using a non-official phrase encoding sheme from 快趣倉頡輸入法.
There were discussions about it on the 倉頡之友 forum, and many Cangjie enthusiasts prefer this method because it applies similar encoding rules between characters and phrases.

Like treating the internal structure of a Chinese character, a phrase is spit apart into 2 or 3 groups of characters, namely head, (body,) tail.
Attached image shows a few examples.

Understood the encoding rules, we can implement them in the Rime dictionary as follows.
A for the first character, B for the second, etc. Z for the last character. Lowercase a, b, z refer to the first, second and last Cangjie letter of the character.

encoder:
  rules:
    - length_equal: 2
      formula: "AaAzBaBbBz"
    - length_equal: 3
      formula: "AaAzBaBzCz"
    - length_in_range: [4, 10]
      formula: "AaBzCaYzZz"

[erikli]

This is so helpful, thank you!