Refactor and clean up the status file parsing code.

tools/testing/dart/status_file.dart

Index: tools/testing/dart/status_file.dart
diff --git a/tools/testing/dart/status_file.dart b/tools/testing/dart/status_file.dart
new file mode 100644
index 0000000000000000000000000000000000000000..cd2d65f2f9511fadf5a356abb930d42fae257000
--- /dev/null
+++ b/tools/testing/dart/status_file.dart
@@ -0,0 +1,164 @@
+// Copyright (c) 2017, the Dart project authors.  Please see the AUTHORS file
+// for details. All rights reserved. Use of this source code is governed by a
+// BSD-style license that can be found in the LICENSE file.
+
+import 'dart:io';
+
+import 'expectation.dart';
+import 'status_expression.dart';
+
+/// Splits out a trailing line comment
+final _commentPattern = new RegExp("^([^#]*)(#.*)?\$");
+
+/// Matches the header that begins a new section, like:
+///
+///     [ $compiler == dart2js && $minified ]
+final _sectionPattern = new RegExp(r"^\[([^\]]+)\]");
+
+/// Matches an entry that defines the status for a path in the current section,
+/// like:
+///
+///     some/path/to/some_test: Pass || Fail
+final _entryPattern = new RegExp(r"\s*([^: ]*)\s*:(.*)");
+
+/// Matches an issue number in a comment, like:
+///
+///     blah_test: Fail # Issue 1234
+///                       ^^^^
+final _issuePattern = new RegExp("[Ii]ssue ([0-9]+)");
+
+/// A parsed status file, which describes how a collection of tests are
+/// expected to behave under various configurations and conditions.
+///
+/// Each status file is made of a series of sections. Each section begins with
+/// a header, followed by a series of entries. A header is enclosed in square
+/// brackets and contains a Boolean expression. That expression is evaluated in
+/// an environment. If it evaluates to true, then the entries after the header
+/// take effect.
+///
+/// Each entry is a glob-like file path followed by a colon and then a
+/// comma-separated list of [Expectation]s. The path may point to an individual
+/// file, or a directory, in which case it applies to all files under that path.
+///
+/// Entries may also appear before any section header, in which case they
+/// always apply.
+class StatusFile {
+  final List<StatusSection> sections = [];
+
+  /// Parses the status file at [path].
+  StatusFile.read(String path) {
+    var lines = new File(path).readAsLinesSync();
+
+    // The current section whose rules are being parsed.
+    StatusSection section;
+
+    var lineNumber = 0;
+    for (var line in lines) {
+      lineNumber++;
+
+      // Strip off the comment and whitespace.
+      var match = _commentPattern.firstMatch(line);
+      var source = "";
+      var comment = "";
+      if (match != null) {
+        source = match[1].trim();
+        comment = match[2] ?? "";
+      }
+
+      // Ignore empty (or comment-only) lines.
+      if (source.isEmpty) continue;
+
+      // See if we are starting a new section.
+      match = _sectionPattern.firstMatch(source);
+      if (match != null) {
+        var condition = Expression.parse(match[1].trim());
+        section = new StatusSection(condition);
+        sections.add(section);
+        continue;
+      }
+
+      // Otherwise, it should be a new entry under the current section.
+      match = _entryPattern.firstMatch(source);
+      if (match != null) {
+        var path = match[1].trim();
+        // TODO(whesse): Handle test names ending in a wildcard (*).
+        var expectations = _parseExpectations(match[2]);
+        var issue = _issueNumber(comment);
+
+        // If we haven't found a section header yet, create an implicit section
+        // that matches everything.
+        if (section == null) {
+          section = new StatusSection(null);
+          sections.add(section);
+        }
+
+        section.entries.add(new StatusEntry(path, expectations, issue));
+        continue;
+      }
+
+      throw new FormatException(
+          "Could not parse line $lineNumber of status file '$path':\n$line");
+    }
+  }
+
+  /// Parses a comma-separated list of expectation names from [text].
+  List<Expectation> _parseExpectations(String text) {

[Bill Hesse, 2017/05/15 16:15:09]

We do a lot of work in ExpectationSet to produce a single cached object for
identical sets.  We could also do something here to produce a single cached list
for identical lines, or use the cache in ExpectationSet (it exists by the time
this is called, but isn't passed in yet).

But this may not be worth it, because it is bounded by the size of the status
files, while the later results (Set<Expectation> from ExpectationSet) are stored
in TestCase objects, one for each test, and they live until the end of the
program.  The StatusFile objects created here may be dropped once the test
suites are finished creating all the test cases, so we don't need to optimize
their memory use as much.

This also applies to the paths in these lines, which are optimized to cached
regexps in ExpectationSet, but can be expected to be distinct here.

+    return text
+        .split(",")
+        .map((name) => Expectation.find(name.trim()))
+        .toList();
+  }
+
+  /// Returns the issue number embedded in [comment] or `null` if there is none.
+  int _issueNumber(String comment) {
+    var match = _issuePattern.firstMatch(comment);
+    if (match == null) return null;
+
+    return int.parse(match[1], onError: (_) => null);
+  }
+
+  String toString() {
+    var buffer = new StringBuffer();
+    for (var section in sections) {
+      buffer.writeln("[${section._condition}]");
+
+      for (var entry in section.entries) {
+        buffer.write("${entry.path}: ${entry.expectations.join(', ')}");
+        if (entry.issue != null) buffer.write(" # Issue ${entry.issue}");
+        buffer.writeln();
+      }
+
+      buffer.writeln();
+    }
+
+    return buffer.toString();
+  }
+}
+
+/// One section in a status file.
+///
+/// Contains the condition from the header that begins the section, then all of
+/// the entries within the section.
+class StatusSection {
+  /// The expression that determines when this section is applied.
+  ///
+  /// May be `null` for paths that appear before any section header in the file.
+  /// In that case, the section always applies.
+  final Expression _condition;
+  final List<StatusEntry> entries = [];
+
+  /// Returns true if this section should apply in the given [environment].
+  bool isEnabled(Map<String, dynamic> environment) =>
+      _condition == null || _condition.evaluate(environment);
+
+  StatusSection(this._condition);
+}
+
+/// Describes the test status of the file or files at a given path.
+class StatusEntry {
+  final String path;
+  final List<Expectation> expectations;
+  final int issue;
+
+  StatusEntry(this.path, this.expectations, this.issue);
+}