How English Input Mode Integrates with Chinese Input in Rime Wanxiang

The wanxiang schema achieves seamless bilingual typing by routing English input through a dedicated schema while the super_english.lua filter formats and spaces text before it re-enters the Chinese processing pipeline.

The Rime wanxiang input method allows users to compose mixed-language sentences without manual mode switching or application changes. This capability stems from a three-layer integration architecture defined across wanxiang.schema.yaml, wanxiang_english.schema.yaml, and lua/super_english.lua. The system treats English not as an isolated mode but as a stream that temporarily bypasses Chinese segmentation yet returns to the same candidate processing pipeline.

Scheme-Level Language Switching

The foundation of bilingual input rests on the ascii_mode switch declared in wanxiang.schema.yaml:

switches:
  - name: ascii_mode          # 中英输入状态

    states: [ 中文, 英文 ]

When activated (typically via Shift+Space as bound in the key_binder section), this switch instructs the Rime engine to treat incoming keystrokes as raw ASCII rather than pinyin. The super_english.lua filter reads this state during its init() function to load English-specific configurations:

env.english_spacing_mode = cfg:get_string("wanxiang_english/english_spacing") or "off"
env.spacing_timeout       = cfg:get_double("wanxiang_english/spacing_timeout") or 0

This initialization ensures that spacing rules and formatting options activate only when ascii_mode is set to "英文".

The English Schema and Dictionary Layer

The dedicated wanxiang_english.schema.yaml defines the English processing environment and references the flat dictionary file wanxiang_english.dict.yaml. This schema configures:

wanxiang_english:
  dictionary: wanxiang_english
  enable_completion: true
  enable_sentence: true
  english_spacing: smart
  spacing_timeout: 5

Setting enable_sentence: true allows the super_english filter to treat complete English phrases as single candidates rather than forcing per-word completion. The dictionary provides raw vocabulary, while the schema parameters control how English text behaves when mixed with Chinese input.

The Super-English Filter: Bridging Both Languages

The lua/super_english.lua file functions as a Lua filter positioned early in the processing chain—after segmentors but before Chinese-specific filters like super_lookup and super_comment. This positioning allows it to format English text while preserving downstream Chinese processing capabilities.

Detecting English Candidates

The filter identifies English input by checking is_ascii_phrase_fast(text) and verifying the presence of letters via has_letters. When ascii_mode is active, incoming ASCII characters route through the English dictionary first, producing raw candidates that super_english intercepts before they reach Chinese-oriented filters.

Smart Spacing and Capitalization

The filter maintains state across commits using prev_commit_is_eng and timestamp tracking (last_commit_time). In smart spacing mode (the default), it automatically inserts spaces between consecutive English words while respecting the spacing_timeout threshold:

if mode == "smart" then
    if code_ctx.prev_is_eng then
        if not find(text, "^%s") then text = " " .. text end
    end
end

The apply_segment_formatting function handles capitalization based on raw input codes, preserving user intent when typing mixed-case strings like "hELLo". The filter also normalizes non-ASCII spaces (such as NBSP) to standard spaces.

Preserving the Chinese Processing Pipeline

After formatting English segments, the filter returns candidates to the pipeline where Chinese-specific processors can still operate. For mixed input like "你好hello", the Chinese segment "你好" passes through unchanged while "hello" receives English formatting. Both segments then continue to filters such as super_lookup (for reverse lookup) and super_comment (for tone hints), which may append metadata without altering the formatted English text.

The filter maintains a short sticky buffer (STICKY_BUFFER_SIZE = 2) to handle backspacing across language boundaries, ensuring that deleting the English portion of a mixed candidate does not destroy the Chinese context.

Configuration and Usage Examples

Toggling Input Modes

Bind the mode switch in wanxiang.schema.yaml:

key_binder:
  bindings:
    - { when: always, accept: Shift+space, select: .next }

Pressing Shift+Space cycles to the next schema state, toggling ascii_mode between Chinese and English.

Mixed Input Behavior

Input sequence:

1. Type "你好" in Chinese mode
2. Toggle to English mode
3. Type "hello world"

Result: "你好 hello world" with automatic spacing inserted by super_english.lua only between the English words.

Timeout-Controlled Spacing

With spacing_timeout: 5, pausing longer than 5 seconds between English words resets the prev_commit_is_eng flag. Subsequent English input will not receive automatic leading spaces, preventing unwanted spacing when resuming typing after a break.

Summary

• Scheme switching: The ascii_mode switch in wanxiang.schema.yaml toggles between Chinese and English input states, triggering the ASCII composer when active.
• Dedicated English layer: wanxiang_english.schema.yaml and its dictionary provide isolated vocabulary and configuration options for English candidates.
• Integration filter: lua/super_english.lua detects English text, applies smart spacing and capitalization, and reinserts formatted candidates into the Chinese processing pipeline.
• State management: The filter tracks previous commits and timing to enable intelligent spacing while maintaining a sticky buffer for seamless backspacing across language boundaries.
• Pipeline compatibility: English formatting occurs before Chinese filters like super_lookup and super_comment, ensuring mixed sentences receive appropriate annotations without formatting loss.

Frequently Asked Questions

How do I toggle between English and Chinese input modes?

Press Shift+Space (or your configured key_binder shortcut) to cycle the ascii_mode switch defined in wanxiang.schema.yaml. This changes the input stream from pinyin to raw ASCII and activates the English formatting filters.

What is the "smart spacing" feature in mixed input?

Smart spacing automatically inserts a single space between consecutive English words while suppressing spaces after Chinese characters or punctuation. The super_english.lua filter tracks prev_commit_is_eng and respects the spacing_timeout setting (default 5 seconds) to determine when to insert these automatic spaces.

Can I customize the English vocabulary in wanxiang?

Yes. Edit wanxiang_english.dict.yaml to add or remove words from the flat dictionary list. The wanxiang_english.schema.yaml controls completion behavior and spacing rules, allowing you to tune how aggressively the system suggests English words alongside Chinese input.

Why does my English text lose formatting when mixed with Chinese?

This typically occurs if the super_english filter is mispositioned in the filters list of wanxiang.schema.yaml. The filter must run before Chinese-specific filters like super_lookup but after segmentors. Verify that lua_filter@super_english appears in the correct sequence to ensure English formatting persists through the entire pipeline.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}