Customizing Input Content

If you want to customize some content, this article provides some explanations.

It is recommended to use the following software to customize the content of this article:

• Visual Studio Code: https://code.visualstudio.com

This article refers to the following sources:

• Wusong Pinyin Documentation: https://dvel.me/posts/rime-ice
• Official Rime Customization Guide: https://github.com/rime/home/wiki/CustomizationGuide

Writing a Lexicon

Due to the design of Rime, English words and super simple abbreviations are not suitable in the Pinyin lexicon:

hello	hello
世界	s j
蒙奇·D·路飞	meng qi d lu fei

As you can see, 世界 is represented by s and j, which will cause the input method to fail to suggest words or phrases starting with s. The same applies to j and d, which will result in the failure to suggest relevant words or phrases starting with j.

If all lexicons are designed in this way, when you input s, the input method will try to suggest all words or phrases starting with s, leading to lagging, memory leaks, and even crashes.

Therefore:

• It is recommended to use full Pinyin for all lexicon entries.
• English lexicons should be placed in the English dictionary. In the Mint input method, the English lexicon is used as the secondary input source, and automatic prediction and sentence generation for English input should be disabled.

When you examine the files in the Mint input method, you will find that the directory structure is as follows:

dicts
├── custom_simple.dict.yaml        # Custom dictionary, where you can add your favorite words
├── luna_pinyin.biaoqing.dict.yaml # Emoji dictionary
├── luna_pinyin.emoji.dict.yaml    # Emoji dictionary (may be removed in the future)
├── luna_pinyin.extended.dict.yaml # Extended dictionary for Luna Pinyin (may be removed in the future)
├── rime_ice.41448.dict.yaml       # Extended single-character dictionary for Rime Ice
├── rime_ice.8105.dict.yaml        # Basic single-character dictionary for Rime Ice
├── rime_ice.base.dict.yaml        # Core basic lexicon for Rime Ice
├── rime_ice.en.dict.yaml          # English dictionary for Rime Ice
├── se_words.dict.yaml             # Commonly used lexicon for software industry
├── terra_rime_ice.base.dict.yaml  # Earth Pinyin dictionary (generated by Python)
└── wubi98_base.dict.yaml          # Wubi 98 dictionary

The content of the lexicon files is written as follows:

---
name: Lexicon Name
version: "Version Number"
sort: by_weight (sort by weight) | original (sort in the order of the character table)
columns:    # When columns attribute is not specified, the default order is:
- text    # Vocabulary
- code    # Code
- weight  # Weight
- stem    # Word-building code (not related to Pinyin scheme)
  ...
  你好	ni hao	123
  For lexicon files without phonetic annotation but with weight specified, modify the columns accordingly:

---

For example, note that the format inside the lexicon is 'Word'<Tab>'Pinyin'<Space>'Pinyin'<Space>'Pinyin'<Tab>'Weight':

# Rime dictionary
# encoding: utf-8
#
# Personalized terms - by @Mintimate
# It is recommended to add custom phrases or words here
---
name: custom_simple
version: "2023.11.30"
sort: by_weight
...

# Personal names
# Common phrases
哈哈	ha ha	99
macOS	mac	99
可以	ke yi	99
# (｡>ㅅ<｡)
Mintimate	mintimate	1
https://www.mintimate.cn	mintimate	2
Mintimate's Blog	mintimate	3

These lexicons are referenced by the lexicon driver configuration in the root directory:

├── custom_dict_en.all.dict.yaml        # English lexicon for Mint input method
├── custom_dict_terra.all.dict.yaml     # Terra Pinyin Mint custom lexicon
├── custom_dict.all.dict.yaml           # Mint Pinyin lexicon
└── custom_dict.wubi.dict.yaml          # Wubi 98 Mint custom lexicon

Let's see how internal references are used:

---
name: custom_dict.all ## Note that the name should match the filename
version: "2020.6.7"
sort: by_weight
# This is where you specify the dictionaries used by the input method, to supplement the extended dictionary.
import_tables:
  - dicts/rime_ice.8105 # Mist Pinyin common character collection
  - dicts/rime_ice.41448 # Mist Pinyin complete character collection
  - dicts/custom_simple # Custom
  - dicts/rime_ice.base # Mist Pinyin https://github.com/iDvel/rime-ice
  - dicts/se_words # Internet network vocabulary
  - dicts/luna_pinyin.biaoqing # Emoticons
  - dicts/luna_pinyin.emoji # Emoji Ext
...

Important：

• name: The name is the filename without the .dict extension, and the filename should end with .dict.
• import_tables: Enumerate the lexicons that need to be imported.

After modifying the lexicon, remember to redeploy the input method.

The above information can help you customize the lexicons.

Custom Text

"Custom Text" refers to the custom_phrase.txt file within the input method. You may not see it in the Mint input method...

In my understanding, "Custom Text" refers to lexicons with particularly high weights (this is the default behavior, but the weights of each translator can be adjusted using initial_quality). Therefore, I have removed the configuration for "Custom Text." If needed, you can configure it yourself. The format is the same as the lexicon:

# Rime table
# encoding: utf-8
# Custom Text
# Do not write any comments after this line
噷	hm
哼	hng


去	q	2
千	q	1

我	w	3
万	w	2
往	w	1

等等	dd
的地得	ddd
等等等等	dddd
刚刚	gg
才刚刚	cgg
知道	zd
不知道	bzd

Also, the input method configuration needs to include:

translators:
    - table_translator@custom_phrase      # Custom Phrase custom_phrase.txt

# Custom Phrase
custom_phrase:
  dictionary: ""
  user_dict: custom_phrase # Need to manually create custom_phrase.txt file
  db_class: stabledb
  enable_completion: false # Completion prompt
  enable_sentence: false # Disable sentence making
  initial_quality: 99    # The weight of custom_phrase should be larger than other lexicon

Custom Text does not interact with other translators in word-building. If you use a complete code, the character or word cannot participate in word-building. That is, self-built words cannot be remembered.

Therefore, it is recommended to fix non-complete code characters or words. For example, '的'(de) should be 'd', '是'(shi) should be 's', and '仙剑'(xian jian) should be 'xj'.

Note that the full Pinyin 'a o e' is also a complete spelling, so single characters of 'a o e' should not be included in the Custom Text. Otherwise, words like '啊 哦 呃' cannot be used for word-building.

Using the full Pinyin input method within Oh-my-rime as an example, according to the Configuration and Overrides method, we can create the rime_mint.custom.yaml file:

# Rime Custom
# encoding: utf-8

patch:
  "engine/translators/+":
    - table_translator@mint_simple            # Mint custom phrases

  # Mint custom phrases
  mint_simple:
    dictionary: ""
    user_dict: dicts/rime_mint.simple
    db_class: stabledb
    enable_completion: false
    enable_sentence: false
    initial_quality: 0.5
    comment_format:
      - xform/^.+$//

At the same time, we need to create the dicts/rime_mint.simple.txt file:

# Rime table
# coding: utf-8
#@/db_name rime_mint.simple.txt
#@/db_type tabledb
#
#
# No comments can be written after this line
我们 wm

Note that the content within the rime_mint.simple.txt file follows the format of 「phrase」<Tab>「Pinyin abbreviation」.

Finally, after redeploying the input method, you will be able to see the effect of the custom text.

Double Pinyin Convert

By default, the Mint input method performs a conversion for the candidate area code in double pinyin. For example, when you need to spell "你好" (Hello) in DoubleFly pinyin, it will appear as "nihao" instead of "nihc".

Actually, this is due to the translator/preedit_format configuration in the scheme. This configuration is used to escape the code. Let's take DoubleFly pinyin as an example:

translator:
  preedit_format:
    - xform/([bpmfdtnljqx])n/$1iao/
    - xform/(\w)g/$1eng/
    - xform/(\w)q/$1iu/
    - xform/(\w)w/$1ei/
    - xform/([dtnlgkhjqxyvuirzcs])r/$1uan/
    - xform/(\w)t/$1ve/
    - xform/(\w)y/$1un/
    - xform/([dtnlgkhvuirzcs])o/$1uo/
    - xform/(\w)p/$1ie/
    - xform/([jqx])s/$1iong/
    - xform/(\w)s/$1ong/
    - xform/(\w)d/$1ai/
    - xform/(\w)f/$1en/
    - xform/(\w)h/$1ang/
    - xform/(\w)j/$1an/
    - xform/([gkhvuirzcs])k/$1uai/
    - xform/(\w)k/$1ing/
    - xform/([jqxnl])l/$1iang/
    - xform/(\w)l/$1uang/
    - xform/(\w)z/$1ou/
    - xform/([gkhvuirzcs])x/$1ua/
    - xform/(\w)x/$1ia/
    - xform/(\w)c/$1ao/
    - xform/([dtgkhvuirzcs])v/$1ui/
    - xform/(\w)b/$1in/
    - xform/(\w)m/$1ian/
    - xform/([aoe])\1(\w)/$1$2/
    - "xform/(^|[ '])v/$1zh/"
    - "xform/(^|[ '])i/$1ch/"
    - "xform/(^|[ '])u/$1sh/"
    - xform/([jqxy])v/$1u/
    - xform/([nl])v/$1ü/
    - xform/ü/v/  # ü 显示为 v

If you don't need it, you can overwrite the translator/preedit_format configuration in the scheme to be empty. For example, in the case of DoubleFly pinyin, we can create a double_pinyin_flypy.custom.yaml file:

# Rime Custom
# encoding: utf-8

patch:
  translator/preedit_format: []

After that, redeploy the input method, and you can see the double pinyin code.

Note

In a custom file, there can only be one patch entry. For example, if I have overwritten other configurations, then the custom file might look like this:

# Rime Custom
# encoding: utf-8

patch:
  "switches/@last":
      name: emoji_suggestion
      reset: 1
      states: [ "😣️","😁️"]
  "engine/translators/+":
    - table_translator@wubi86_jidian
  wubi86_jidian:
    dictionary: wubi86_jidian           # English dictionary
    enable_sentence: false         # Turn off automatic sentence making
    enable_completion: false       # Turn off automatic prompts
    initial_quality: 0.8
  "engine/filters/+":
    - lua_filter@*tag_user_dict               # Mark user's phrases and dictionaries
  # Dictionary prompts
  tag_user_dict:
    # User dictionary representation
    user_table: '☁'
    # Auto full
    completion: '☁'
    # Auto sentence making
    sentence: '~'
    # Default phrase
    phrase: ''
    # User phrase
    user_phrase: '*'
  # Code escape
  translator/preedit_format: []