Support custom spoken forms for graphemes

cursorless-talon/src/get_grapheme_spoken_form_entries.py

@@ -0,0 +1,75 @@
+import re
+import typing
+from collections import defaultdict
+from typing import Iterator, Mapping
+from uu import Error
+
+from talon import app, registry
+
+from .spoken_forms_output import SpokenFormOutputEntry
+
+grapheme_capture_name = "user.any_alphanumeric_key"
+
+
+def get_grapheme_spoken_form_entries() -> list[SpokenFormOutputEntry]:
+    return [
+        {
+            "type": "grapheme",
+            "id": id,
+            "spokenForms": spoken_forms,
+        }
+        for symbol_list in generate_lists_from_capture(grapheme_capture_name)
+        for id, spoken_forms in get_id_to_spoken_form_map(symbol_list).items()
+    ]
+
+
+def generate_lists_from_capture(capture_name) -> Iterator[str]:
+    """
+    Given the name of a capture, yield the names of each list that the capture
+    expands to. Note that we are somewhat strict about the format of the
+    capture rule, and will not handle all possible cases.
+    """
+    if capture_name.startswith("self."):
+        capture_name = "user." + capture_name[5:]
+    try:
+        rule = registry.captures[capture_name][0].rule.rule
+    except Error:
+        app.notify("Error constructing spoken forms for graphemes")
+        print(f"Error getting rule for capture {capture_name}")
+        return
+    rule = rule.strip()
+    if rule.startswith("(") and rule.endswith(")"):
+        rule = rule[1:-1]
+        rule = rule.strip()
+    components = re.split(r"\s*\|\s*", rule)
+    for component in components:
+        if component.startswith("<") and component.endswith(">"):
+            yield from generate_lists_from_capture(component[1:-1])
+        elif component.startswith("{") and component.endswith("}"):
+            component = component[1:-1]
+            if component.startswith("self."):
+                component = "user." + component[5:]
+            yield component
+        else:
+            app.notify("Error constructing spoken forms for graphemes")
+            print(
+                f"Unexpected component {component} while processing rule {rule} for capture {capture_name}"
+            )
+
+
+def get_id_to_spoken_form_map(list_name: str) -> Mapping[str, list[str]]:
+    """
+    Given the name of a Talon list, return a mapping from the values in that
+    list to the list of spoken forms that map to the given value.
+    """
+    try:
+        raw_list = typing.cast(dict[str, str], registry.lists[list_name][0]).copy()
+    except Error:
+        app.notify(f"Error getting list {list_name}")
+        return {}
+
+    inverted_list: defaultdict[str, list[str]] = defaultdict(list)
+    for key, value in raw_list.items():
+        inverted_list[value].append(key)
+
+    return inverted_list

cursorless-talon/src/spoken_forms.py

@@ -2,7 +2,7 @@
 from pathlib import Path
 from typing import Callable, Concatenate, ParamSpec, TypeVar
 
-from talon import app, fs
+from talon import app, cron, fs, registry
 
 from .actions.actions import ACTION_LIST_NAMES
 from .csv_overrides import (
@@ -11,6 +11,10 @@
     SpokenFormEntry,
     init_csv_and_watch_changes,
 )
+from .get_grapheme_spoken_form_entries import (
+    get_grapheme_spoken_form_entries,
+    grapheme_capture_name,
+)
 from .marks.decorated_mark import init_hats
 from .spoken_forms_output import SpokenFormsOutput
 
@@ -99,14 +103,17 @@ def update():
     def update_spoken_forms_output():
         spoken_forms_output.write(
             [
-                {
-                    "type": LIST_TO_TYPE_MAP[entry.list_name],
-                    "id": entry.id,
-                    "spokenForms": entry.spoken_forms,
-                }
-                for spoken_form_list in custom_spoken_forms.values()
-                for entry in spoken_form_list
-                if entry.list_name in LIST_TO_TYPE_MAP
+                *[
+                    {
+                        "type": LIST_TO_TYPE_MAP[entry.list_name],
+                        "id": entry.id,
+                        "spokenForms": entry.spoken_forms,
+                    }
+                    for spoken_form_list in custom_spoken_forms.values()
+                    for entry in spoken_form_list
+                    if entry.list_name in LIST_TO_TYPE_MAP
+                ],
+                *get_grapheme_spoken_form_entries(),
             ]
         )
 
@@ -184,9 +191,30 @@ def on_watch(path, flags):
         update()
 
 
+update_captures_cron = None
+
+
+def update_captures_debounced(updated_captures: set[str]):
+    if grapheme_capture_name not in updated_captures:
+        return
+
+    global update_captures_cron
+    cron.cancel(update_captures_cron)
+    update_captures_cron = cron.after("100ms", update_captures)
+
+
+def update_captures():
+    global update_captures_cron
+    update_captures_cron = None
+
+    update()
+
+
 def on_ready():
     update()
 
+    registry.register("update_captures", update_captures_debounced)
+
     fs.watch(str(JSON_FILE.parent), on_watch)
 
 

cursorless-talon/src/spoken_forms_output.py

@@ -8,7 +8,7 @@
 STATE_JSON_VERSION_NUMBER = 0
 
 
-class SpokenFormEntry(TypedDict):
+class SpokenFormOutputEntry(TypedDict):
     type: str
     id: str
     spokenForms: list[str]
@@ -29,7 +29,7 @@ def init(self):
             print(error_message)
             app.notify(error_message)
 
-    def write(self, spoken_forms: list[SpokenFormEntry]):
+    def write(self, spoken_forms: list[SpokenFormOutputEntry]):
         with open(SPOKEN_FORMS_OUTPUT_PATH, "w", encoding="UTF-8") as out:
             try:
                 out.write(

packages/cursorless-engine/src/generateSpokenForm/CustomSpokenFormGeneratorImpl.test.ts

@@ -19,6 +19,11 @@ suite("CustomSpokenFormGeneratorImpl", async function () {
               id: "setSelection",
               spokenForms: ["bar"],
             },
+            {
+              type: "grapheme",
+              id: "a",
+              spokenForms: ["alabaster"],
+            },
           ];
         },
         onDidChange: () => ({ dispose() {} }),
@@ -33,7 +38,7 @@ suite("CustomSpokenFormGeneratorImpl", async function () {
         }),
         {
           type: "success",
-          spokenForms: ["foo air"],
+          spokenForms: ["foo alabaster"],
         },
       );
       assert.deepStrictEqual(

packages/cursorless-engine/src/generateSpokenForm/primitiveTargetToSpokenForm.ts

@@ -20,7 +20,6 @@ import {
   numberToSpokenForm,
   ordinalToSpokenForm,
 } from "./defaultSpokenForms/numbers";
-import { characterToSpokenForm } from "./defaultSpokenForms/characters";
 import { SpokenFormComponentMap } from "./getSpokenFormComponentMap";
 import { SpokenFormComponent } from "./SpokenFormComponent";
 
@@ -234,7 +233,7 @@ export class PrimitiveTargetSpokenFormGenerator {
       case "glyph":
         return [
           this.spokenFormMap.complexScopeTypeType.glyph,
-          characterToSpokenForm(scopeType.character),
+          this.spokenFormMap.grapheme[scopeType.character],
         ];
       case "surroundingPair": {
         const pair = this.spokenFormMap.pairedDelimiter[scopeType.delimiter];
@@ -274,14 +273,14 @@ export class PrimitiveTargetSpokenFormGenerator {
     switch (mark.type) {
       case "decoratedSymbol": {
         const [color, shape] = mark.symbolColor.split("-");
-        const components: string[] = [];
+        const components: SpokenFormComponent[] = [];
         if (color !== "default") {
           components.push(hatColorToSpokenForm(color));
         }
         if (shape != null) {
           components.push(hatShapeToSpokenForm(shape));
         }
-        components.push(characterToSpokenForm(mark.character));
+        components.push(this.spokenFormMap.grapheme[mark.character]);
         return components;
       }
 

packages/cursorless-engine/src/scopeProviders/TalonSpokenForms.ts

@@ -24,6 +24,7 @@ export const SUPPORTED_ENTRY_TYPES = [
   "pairedDelimiter",
   "action",
   "customAction",
+  "grapheme",
 ] as const;
 
 type SupportedEntryType = (typeof SUPPORTED_ENTRY_TYPES)[number];

packages/cursorless-engine/src/spokenForms/SpokenFormType.ts

@@ -42,6 +42,11 @@ export interface SpokenFormMapKeyTypes {
    * custom actions corresponding to id's of VSCode commands.
    */
   customAction: string;
+
+  /**
+   * Individual characters / graphemes, eg `a` or `/`.
+   */
+  grapheme: string;
 }
 
 /**

packages/cursorless-engine/src/spokenForms/defaultSpokenFormMapCore.ts

@@ -1,4 +1,5 @@
 import { DefaultSpokenFormMapDefinition } from "./defaultSpokenFormMap.types";
+import { graphemeDefaultSpokenForms } from "./graphemes";
 import { isDisabledByDefault, isPrivate } from "./spokenFormMapUtil";
 
 /**
@@ -211,4 +212,5 @@ export const defaultSpokenFormMapCore: DefaultSpokenFormMapDefinition = {
     // nextHomophone: "phones",
   },
   customAction: {},
+  grapheme: graphemeDefaultSpokenForms,
 };

packages/cursorless-engine/src/generateSpokenForm/defaultSpokenForms/characters.ts → packages/cursorless-engine/src/spokenForms/graphemes.ts

@@ -1,7 +1,5 @@
 /* eslint-disable @typescript-eslint/naming-convention */
 
-import { NoSpokenFormError } from "../NoSpokenFormError";
-
 // https://github.com/talonhub/community/blob/9acb6c9659bb0c9b794a7b7126d025603b4ed726/core/keys/keys.py
 
 const alphabet = Object.fromEntries(
@@ -59,16 +57,8 @@ const symbols = {
   "\uFFFD": "special",
 };
 
-const characters: Record<string, string> = {
+export const graphemeDefaultSpokenForms: Record<string, string> = {
   ...alphabet,
   ...digits,
   ...symbols,
 };
-
-export function characterToSpokenForm(char: string): string {
-  const result = characters[char];
-  if (result == null) {
-    throw new NoSpokenFormError(`Unknown character '${char}'`);
-  }
-  return result;
-}