Fix autocomplete and suggestions in the analyzer plugin

extras/plugin_example/README.md

@@ -4,7 +4,9 @@ Currently, we support
 
 - showing errors in moor files
 - outline
-- (kinda) folding
+- folding
+- (very, very limited) autocomplete
+-  some quickfixes to make columns nullable or non-null
 
 ## Setup
 To use this plugin, you'll need to perform these steps once. It is assumed that you

extras/plugin_example/analysis_options.yaml

@@ -1,5 +1,5 @@
 include: package:pedantic/analysis_options.yaml
 
-#analyzer:
-#  plugins:
-#    - moor
+analyzer:
+  plugins:
+    - moor

moor_generator/lib/src/backends/plugin/backend/driver.dart

@@ -80,9 +80,15 @@ class MoorDriver implements AnalysisDriverGeneric {
       }
       final backendTask = _createTask(mostImportantFile.file.uri);
 
-      final task = session.startTask(backendTask);
-      await task.runTask();
-      _tracker.handleTaskCompleted(task);
+      try {
+        final task = session.startTask(backendTask);
+        await task.runTask();
+        _tracker.handleTaskCompleted(task);
+      } catch (e, s) {
+        Logger.root.warning(
+            'Error while working on ${mostImportantFile.file.uri}', e, s);
+        _tracker.removePending(mostImportantFile);
+      }
     } finally {
       _isWorking = false;
     }
@@ -141,11 +147,16 @@ class MoorDriver implements AnalysisDriverGeneric {
 
   /// Waits for the file at [path] to be parsed.
   Future<FoundFile> waitFileParsed(String path) {
-    _scheduler.notify(this);
-
     final found = pathToFoundFile(path);
 
-    return completedFiles()
-        .firstWhere((file) => file == found && file.isParsed);
+    if (found.isParsed) {
+      return Future.value(found);
+    } else {
+      _scheduler.notify(this);
+      final found = pathToFoundFile(path);
+
+      return completedFiles()
+          .firstWhere((file) => file == found && file.isParsed);
+    }
   }
 }

moor_generator/lib/src/backends/plugin/backend/file_tracker.dart

@@ -76,6 +76,13 @@ class FileTracker {
     }
   }
 
+  /// Manually remove the [file] from the backlog. As the plugin is still very
+  /// unstable, we use this on unexpected errors so that we just move on to the
+  /// next file if there is a problem.
+  void removePending(TrackedFile file) {
+    _pendingWork.remove(file);
+  }
+
   void dispose() {
     _computations.close();
   }

moor_generator/lib/src/backends/plugin/services/assists/assist_service.dart

@@ -37,6 +37,12 @@ abstract class _AssistOnNodeContributor<T extends AstNode> {
   const _AssistOnNodeContributor();
 
   void contribute(AssistCollector collector, T node, String path);
+
+  SourceEdit replaceNode(AstNode node, String text) {
+    final start = node.firstPosition;
+    final length = node.lastPosition - start;
+    return SourceEdit(start, length, text);
+  }
 }
 
 class AssistId {

moor_generator/lib/src/backends/plugin/services/assists/column_nullability.dart

@@ -9,9 +9,9 @@ class ColumnNullability extends _AssistOnNodeContributor<ColumnDefinition> {
     final notNull = node.findConstraint<NotNull>();
 
     if (notNull == null) {
-      // there is no not-null constraint on this column, suggest to add one at
-      // the end of the definition
-      final end = node.lastPosition;
+      // there is no not-null constraint on this column, suggest to add one
+      // after the type name
+      final end = node.typeNames.last.span.end.offset;
       final id = AssistId.makeNotNull;
 
       collector.addAssist(PrioritizedSourceChange(
@@ -34,9 +34,7 @@ class ColumnNullability extends _AssistOnNodeContributor<ColumnDefinition> {
       collector.addAssist(PrioritizedSourceChange(
         id.priority,
         SourceChange('Make this column nullable', id: id.id, edits: [
-          SourceFileEdit(path, -1, edits: [
-            SourceEdit(notNull.firstPosition, notNull.lastPosition, '')
-          ])
+          SourceFileEdit(path, -1, edits: [replaceNode(notNull, '')])
         ]),
       ));
     }

sqlparser/lib/src/ast/schema/column_definition.dart

@@ -6,6 +6,9 @@ class ColumnDefinition extends AstNode {
   final String typeName;
   final List<ColumnConstraint> constraints;
 
+  /// The tokens there were involved in defining the type of this column.
+  List<Token> typeNames;
+
   ColumnDefinition(
       {@required this.columnName,
       @required this.typeName,

sqlparser/lib/src/engine/autocomplete/engine.dart

@@ -21,11 +21,19 @@ class AutoCompleteEngine {
   final List<Hint> _hints = [];
   UnmodifiableListView<Hint> _hintsView;
 
+  final List<Token> _tokens;
+
   void addHint(Hint hint) {
-    _hints.insert(_lastHintBefore(hint.offset), hint);
+    if (_hints.isEmpty) {
+      _hints.add(hint);
+    } else {
+      // ensure that the hints list stays sorted by offset
+      final position = _lastHintBefore(hint.offset);
+      _hints.insert(position + 1, hint);
+    }
   }
 
-  AutoCompleteEngine() {
+  AutoCompleteEngine(this._tokens) {
     _hintsView = UnmodifiableListView(_hints);
   }
 
@@ -40,13 +48,23 @@ class AutoCompleteEngine {
     final hint = foundHints[_lastHintBefore(offset)];
 
     final suggestions = hint.description.suggest(CalculationRequest()).toList();
-    return ComputedSuggestions(hint.offset, offset - hint.offset, suggestions);
+
+    // when calculating the offset, respect whitespace that comes after the
+    // last hint.
+    final lastToken = hint.before;
+    final nextToken =
+        lastToken != null ? _tokens[lastToken.index + 1] : _tokens.first;
+
+    final hintOffset = nextToken.span.start.offset;
+    final length = offset - hintOffset;
+
+    return ComputedSuggestions(hintOffset, length, suggestions);
   }
 
+  /// find the last hint that appears before [offset]
   int _lastHintBefore(int offset) {
-    // find the last hint that appears before offset
     var min = 0;
-    var max = foundHints.length;
+    var max = foundHints.length - 1;
 
     while (min < max) {
       final mid = min + ((max - min) >> 1);
@@ -56,10 +74,18 @@ class AutoCompleteEngine {
 
       if (offsetOfMid == offset) {
         return mid;
-      } else if (offsetOfMid < offset) {
-        min = mid + 1;
       } else {
-        max = mid - 1;
+        final offsetOfNext = _hints[mid + 1].offset;
+
+        if (offsetOfMid < offset) {
+          if (offsetOfNext > offset) {
+            // next one is too late, so this must be the correct one
+            return mid;
+          }
+          min = mid + 1;
+        } else {
+          max = mid - 1;
+        }
       }
     }
 

sqlparser/lib/src/engine/sql_engine.dart

@@ -66,8 +66,9 @@ class SqlEngine {
   ParseResult parseMoorFile(String content) {
     assert(useMoorExtensions);
 
-    final autoComplete = AutoCompleteEngine();
     final tokens = tokenize(content);
+    final autoComplete = AutoCompleteEngine(tokens);
+
     final tokensForParser = tokens.where((t) => !t.invisibleToParser).toList();
     final parser =
         Parser(tokensForParser, useMoor: true, autoComplete: autoComplete);

sqlparser/lib/src/reader/parser/schema.dart

@@ -80,7 +80,14 @@ mixin SchemaParser on ParserBase {
     final name = _consume(TokenType.identifier, 'Expected a column name')
         as IdentifierToken;
 
-    final typeName = _typeName();
+    final typeNames = _typeName();
+    if (typeNames == null) {
+      _error('Expected a type name here');
+    }
+
+    final typeSpan = typeNames.first.span.expand(typeNames.last.span);
+    final typeName = typeSpan.text;
+
     final constraints = <ColumnConstraint>[];
     ColumnConstraint constraint;
     while ((constraint = _columnConstraint(orNull: true)) != null) {
@@ -91,19 +98,21 @@ mixin SchemaParser on ParserBase {
       columnName: name.identifier,
       typeName: typeName,
       constraints: constraints,
-    )..setSpan(name, _previous);
+    )
+      ..setSpan(name, _previous)
+      ..typeNames = typeNames;
   }
 
-  String _typeName() {
+  List<Token> _typeName() {
     // sqlite doesn't really define what a type name is and has very loose rules
     // at turning them into a type affinity. We support this pattern:
     // typename = identifier [ "(" { identifier | comma | number_literal } ")" ]
     if (!_matchOne(TokenType.identifier)) return null;
 
-    final typeNameBuilder = StringBuffer(_previous.lexeme);
+    final typeNames = [_previous];
 
     if (_matchOne(TokenType.leftParen)) {
-      typeNameBuilder.write('(');
+      typeNames.add(_previous);
 
       const inBrackets = [
         TokenType.identifier,
@@ -111,14 +120,15 @@ mixin SchemaParser on ParserBase {
         TokenType.numberLiteral
       ];
       while (_match(inBrackets)) {
-        typeNameBuilder..write(' ')..write(_previous.lexeme);
+        typeNames.add(_previous);
       }
 
       _consume(TokenType.rightParen,
           'Expected closing paranthesis to finish type name');
+      typeNames.add(_previous);
     }
 
-    return typeNameBuilder.toString();
+    return typeNames;
   }
 
   ColumnConstraint _columnConstraint({bool orNull = false}) {
@@ -127,10 +137,13 @@ mixin SchemaParser on ParserBase {
     final resolvedName = _constraintNameOrNull();
 
     if (_matchOne(TokenType.primary)) {
+      _suggestHint(HintDescription.token(TokenType.key));
       _consume(TokenType.key, 'Expected KEY to complete PRIMARY KEY clause');
 
       final mode = _orderingModeOrNull();
       final conflict = _conflictClauseOrNull();
+
+      _suggestHint(HintDescription.token(TokenType.autoincrement));
       final hasAutoInc = _matchOne(TokenType.autoincrement);
 
       return PrimaryKeyColumn(resolvedName,
@@ -138,6 +151,8 @@ mixin SchemaParser on ParserBase {
         ..setSpan(first, _previous);
     }
     if (_matchOne(TokenType.not)) {
+      _suggestHint(HintDescription.token(TokenType.$null));
+
       final notToken = _previous;
       final nullToken =
           _consume(TokenType.$null, 'Expected NULL to complete NOT NULL');
@@ -249,6 +264,7 @@ mixin SchemaParser on ParserBase {
   }
 
   ConflictClause _conflictClauseOrNull() {
+    _suggestHint(HintDescription.token(TokenType.on));
     if (_matchOne(TokenType.on)) {
       _consume(TokenType.conflict,
           'Expected CONFLICT to complete ON CONFLICT clause');
@@ -260,6 +276,7 @@ mixin SchemaParser on ParserBase {
         TokenType.ignore: ConflictClause.ignore,
         TokenType.replace: ConflictClause.replace,
       };
+      _suggestHint(HintDescription.tokens(modes.keys.toList()));
 
       if (_match(modes.keys)) {
         return modes[_previous.type];
@@ -284,7 +301,10 @@ mixin SchemaParser on ParserBase {
 
     ReferenceAction onDelete, onUpdate;
 
+    _suggestHint(HintDescription.token(TokenType.on));
     while (_matchOne(TokenType.on)) {
+      _suggestHint(
+          const HintDescription.tokens([TokenType.delete, TokenType.update]));
       if (_matchOne(TokenType.delete)) {
         onDelete = _referenceAction();
       } else if (_matchOne(TokenType.update)) {

sqlparser/lib/src/reader/tokenizer/scanner.dart

@@ -35,6 +35,11 @@ class Scanner {
 
     final endSpan = _file.span(source.length);
     tokens.add(Token(TokenType.eof, endSpan));
+
+    for (var i = 0; i < tokens.length; i++) {
+      tokens[i].index = i;
+    }
+
     return tokens;
   }
 

sqlparser/lib/src/reader/tokenizer/token.dart

@@ -260,7 +260,10 @@ class Token {
   final FileSpan span;
   String get lexeme => span.text;
 
-  const Token(this.type, this.span);
+  /// The index of this [Token] in the list of tokens scanned.
+  int index;
+
+  Token(this.type, this.span);
 
   @override
   String toString() {
@@ -274,7 +277,7 @@ class StringLiteralToken extends Token {
   /// sqlite allows binary strings (x'literal') which are interpreted as blobs.
   final bool binary;
 
-  const StringLiteralToken(this.value, FileSpan span, {this.binary = false})
+  StringLiteralToken(this.value, FileSpan span, {this.binary = false})
       : super(TokenType.stringLiteral, span);
 }
 
@@ -295,7 +298,7 @@ class IdentifierToken extends Token {
     }
   }
 
-  const IdentifierToken(this.escaped, FileSpan span, {this.synthetic = false})
+  IdentifierToken(this.escaped, FileSpan span, {this.synthetic = false})
       : super(TokenType.identifier, span);
 }