From 26444e729104b9f9dfdf8f78fd4ceaad701fd211 Mon Sep 17 00:00:00 2001
From: Joshua Horton <joshua_horton@sil.org>
Date: Fri, 15 Aug 2025 14:04:52 -0500
Subject: [PATCH] fix(web): fix empty context handling

---
 .../worker-thread/src/main/correction/context-state.ts | 10 ++++------
 .../src/main/correction/context-tokenization.ts        |  2 +-
 2 files changed, 5 insertions(+), 7 deletions(-)

diff --git a/web/src/engine/predictive-text/worker-thread/src/main/correction/context-state.ts b/web/src/engine/predictive-text/worker-thread/src/main/correction/context-state.ts
index f9154a27f1..6d9ff73eea 100644
--- a/web/src/engine/predictive-text/worker-thread/src/main/correction/context-state.ts
+++ b/web/src/engine/predictive-text/worker-thread/src/main/correction/context-state.ts
@@ -181,13 +181,11 @@ export class ContextState {
 
     // Apply all transforms to the base context state
     const transformSequenceDistribution = tokenizeAndFilterDistribution(context, lexicalModel, transformDistribution);
+    const postContext = transformDistribution?.[0] ? applyTransform(transformDistribution[0].sample, context) : context;
 
-    if(transformDistribution?.[0]) {
-      context = applyTransform(transformDistribution[0].sample, context);
-    }
     // Note for future:  the next line's pattern asserts that there is only one true tokenization.
     // We may eventually allow for multiple potential tokenizations (per epic-dict-breaker)
-    const tokenizedContext = determineModelTokenizer(lexicalModel)(context).left;
+    const tokenizedContext = determineModelTokenizer(lexicalModel)(postContext).left;
     if(tokenizedContext.length == 0) {
       tokenizedContext.push({text: ''});
     }
@@ -203,7 +201,7 @@ export class ContextState {
     }
 
     const resultTokenization = this.tokenization.transitionTo(
-      determineModelTokenizer(lexicalModel)(context).left,
+      tokenizedContext,
       alignmentResults,
       lexicalModel,
       transformSequenceDistribution
@@ -242,7 +240,7 @@ export class ContextState {
       preservationTransform = preservationTransform ? buildMergedTransform(preservationTransform, primaryInput): primaryInput;
     }
 
-    const state = new ContextState(context, lexicalModel);
+    const state = new ContextState(postContext, lexicalModel);
     state.tokenization = resultTokenization;
     state.appliedInput = transformDistribution?.[0].sample;
     transition.finalize(state, transformDistribution, preservationTransform);
diff --git a/web/src/engine/predictive-text/worker-thread/src/main/correction/context-tokenization.ts b/web/src/engine/predictive-text/worker-thread/src/main/correction/context-tokenization.ts
index 3f903357b0..a6ad9757f5 100644
--- a/web/src/engine/predictive-text/worker-thread/src/main/correction/context-tokenization.ts
+++ b/web/src/engine/predictive-text/worker-thread/src/main/correction/context-tokenization.ts
@@ -317,7 +317,7 @@ export class ContextTokenization {
    * with the tokenization represented by the current instance.
    * @param alignment The alignment, as determined by a prior call to `computeAlignment`.
    * @param lexicalModel The active lexical model
-   * @param alignedTransformDistribution The tokenized version of the input Transform instance accounting
+   * @param alignedTransformDistribution The tokenized version of the input distribution accounting
    * for the difference between the context represented by this instance and that of `tokenizedContext`.
    * @returns
    */