Fix completer tests (make token substitution case sensitive, otherwise

selecting `ValueError` on `va` would produce `valueError`.

Author: krassowski

atest/05_Features/Completion.robot

@@ -150,7 +150,7 @@ Continuous Hinting Works
     [Setup]    Prepare File for Editing    Python    completion    completion.py
     Configure JupyterLab Plugin    {"continuousHinting": true}    plugin id=${COMPLETION PLUGIN ID}
     # TODO: remove once we resolve https://github.com/jupyterlab/jupyterlab/issues/15022
-    Configure JupyterLab Plugin    {"autoCompletion": true}    plugin id=${MANAGER PLUGIN ID}
+    Configure JupyterLab Plugin    {"autoCompletion": true, "providerTimeout": 2500}    plugin id=${MANAGER PLUGIN ID}
     Place Cursor In File Editor At    9    2
     Wait For Ready State
     Press Keys    None    d
@@ -373,6 +373,8 @@ Completes Paths In Strings
 *** Keywords ***
 Setup Completion Test
     Setup Notebook    Python    Completion.ipynb
+    # TODO: this should be per-provider (upstream issue)
+    Configure JupyterLab Plugin    {"providerTimeout": 2500}    plugin id=${MANAGER PLUGIN ID}
 
 Get Cell Editor Content
     [Arguments]    ${cell_nr}

packages/jupyterlab-lsp/src/features/completion/model.ts

@@ -65,8 +65,28 @@ export class GenericCompleterModel<
     if (this.current && this.cursor) {
       // set initial query to pre-filter items; in future we should use:
       // https://github.com/jupyterlab/jupyterlab/issues/9763#issuecomment-1001603348
+
+      // note: start/end from cursor are not ideal because these get populated from fetch
+      // reply which will vary depending on what providers decide to return; we want the
+      // actual position in token, the same as passed in request to fetch. We can get it
+      // by searching for longest common prefix as seen below (or by counting characters).
+      // Maybe upstream should expose it directly?
       const { start, end } = this.cursor;
-      let query = this.current.text.substring(start, end).trim();
+      const { text, line, column } = this.original!;
+
+      const queryRange = text.substring(start, end).trim();
+      const linePrefix = text.split('\n')[line].substring(0, column).trim();
+      let query = '';
+      for (let i = queryRange.length; i > 0; i--) {
+        if (queryRange.slice(0, i) == linePrefix.slice(-i)) {
+          query = linePrefix.slice(-i);
+          break;
+        }
+      }
+      if (!query) {
+        return;
+      }
+
       let trimmedQuotes = false;
       // special case for "Completes Paths In Strings" test case
       if (query.startsWith('"') || query.startsWith("'")) {

packages/jupyterlab-lsp/src/features/completion/provider.ts

@@ -343,10 +343,10 @@ export function transformLSPCompletions<T>(
     }
 
     // Add overlap with token prefix
-    if (intendedText.toLowerCase().startsWith(token.value.toLowerCase())) {
+    if (intendedText.startsWith(token.value)) {
       anyPrefixed = true;
       // remove overlap with prefix before expanding it
-      if (intendedText.toLowerCase().startsWith(prefix.toLowerCase())) {
+      if (intendedText.startsWith(prefix)) {
         text = text.substring(prefix.length, text.length);
         match.insertText = text;
       }
@@ -364,7 +364,8 @@ export function transformLSPCompletions<T>(
         text =
           (prefix.startsWith("'") || prefix.startsWith('"') ? prefix[0] : '') +
           pathPrefix +
-          text;
+          text +
+          (suffix.startsWith("'") || suffix.startsWith('"') ? suffix[0] : '');
         match.insertText = text;
         // for label without quotes
         match.label = pathPrefix + match.label;