Fix/MultiFile pattern (Issue #696) by Pr0metheus2 · Pull Request #697 · LogExperts/LogExpert

Pr0metheus2 · 2026-08-26T10:25:34Z

This is a fix for #696

Fixed/MultiFile rollover patterns with an index before the extension:
- RolloverFilenameBuilder: the parser was dropping literal text after the final placeholder, including .log.
- *$J(.).log now correctly expands: app.log → app.1.log → app.2.log.
Updated MultiFile mask dialog legend:
- Fixed incorrectly displayed <date> and <prefix> placeholders in English and German resources.
- Added both usage examples:
  - *$J(.) → app.log, app.log.1, app.log.2
  - *$J(.).log → app.log, app.1.log, app.2.log
- Added the examples in English, German, and Chinese.
- Enlarged the dialog and syntax-help label so the date-format lines are no longer clipped.

Before

After

Fixed/Items in menu File/Multifile -> "Enable MultiFile" and "File name mask...", were always greyed out. Now they are displayed normally after the .log file is open.
- removed the BackColor and ForeColor assignments for both:
  - multiFileEnabledStripMenuItem
  - multifileMaskToolStripMenuItem

Before

After

Added regression coverage:
- Filename-builder tests for app.1.log and app.2.log.
- Reader-level test that loads app.log, app.1.log, and app.2.log in correct order using *$J(.).log.
- Kept the existing *$J(.) coverage for app.log.1 / app.log.2.

Verification:
- Full test suite: 1064 passed, 0 failed, 7 skipped.
- Resource project build: passed with 0 warnings, 0 errors.

- Fixed/MultiFile rollover patterns with an index before the extension: - RolloverFilenameBuilder: the parser was dropping literal text after the final placeholder, including .log. - *$J(.).log now correctly expands: app.log → app.1.log → app.2.log. - Updated MultiFile mask dialog legend: - Fixed incorrectly displayed <date> and <prefix> placeholders in English and German resources. - Added both usage examples: *$J(.) → app.log, app.log.1, app.log.2 *$J(.).log → app.log, app.1.log, app.2.log - Added the examples in English, German, and Chinese. - Enlarged the dialog and syntax-help label so the date-format lines are no longer clipped. - Fixed/Items in menu File/Multifile -> "Enable MultiFile" and "Multi file mask...", were always greyed out. Now they are displayed normally after the .log file is open. - removed the BackColor and ForeColor assignments for both: - multiFileEnabledStripMenuItem - multifileMaskToolStripMenuItem - Added regression coverage: - Filename-builder tests for app.1.log and app.2.log. - Reader-level test that loads app.log, app.1.log, and app.2.log in correct order using *$J(.).log. - Kept the existing *$J(.) coverage for app.log.1 / app.log.2. Verification: - Full test suite: 1064 passed, 0 failed, 7 skipped. - Resource project build: passed with 0 warnings, 0 errors.

Hirogen · 2026-08-28T08:16:01Z

            }
        }

+        _ = result.Append(Regex.Escape(segment.ToString()));


Duplicated Code — [RolloverFilenameBuilder.cs:241] is the exact shape of the state-0 flush at line 202. A tiny local FlushEscaped() would let both share it

Added FlushEscaped() to share duplicated escaping code.

Hirogen · 2026-08-28T08:22:44Z

            }
        }

+        _ = result.Append(Regex.Escape(segment.ToString()));


Index parsing from a rotated file is still broken. Spec: "app.1.log → Rotated log (Index 1)". RolloverFilenameHandler calls SetFileName on whatever file the user opens; for $J(.).log the regex .(?'index'\d*).log has a greedy .* that swallows the digit in app.1.log, so Index parses as 0 and BuildFileName would produce app.1.1.log. Grouping only works when the active file (app.log) is opened. The same flaw pre-exists for *$J(.) with app.log.1, so it's parity rather than a regression — so the index mapping is only partially delivered.

fixed parsing of $J(<prefix>) from rotated files, so both app.log.1 and app.1.log correctly identify index 1.

made wildcard matching non-greedy and anchored the generated regex to the complete filename.

updated rebuilding logic to remove an existing prefix plus index together before inserting the new index.

fixed $D(...) parsing to use positions in the escaped format string, preventing errors when a literal regex character such as . appears before $D

Added coverage for opening:

engine.log.1 with *$J(.) → builds engine.log.2

engine.1.log with *$J(.).log → builds engine.2.log

Hirogen · 2026-08-28T08:23:37Z

    [TestCase("engine1.log", "engine2.log","engine$J.log")]
    [TestCase("engine.log", "engine.log.1","*$J(.)")]
+    [TestCase("engine.log", "engine.1.log", "*$J(.).log")]
    [TestCase("engine_2010-06-12.log", "engine_2010-06-12.log.1", "*$D(yyyy-MM-dd).log$J(.)")]


$I untested. For "$J, $I" support, the fix itself generalizes, but no test covers $I or $D followed by a trailing literal (e.g. *$D(yyyy-MM-dd)$J(.).log).

Added cases which cover:

$I with a trailing literal .log, starting from a rotated file.

$D plus $J(.) followed by a trailing .log, also starting from a rotated file.

Hirogen · 2026-08-28T08:24:49Z

    [TestCase("engine1.log", "engine2.log","engine$J.log")]
    [TestCase("engine.log", "engine.log.1","*$J(.)")]
+    [TestCase("engine.log", "engine.1.log", "*$J(.).log")]
    [TestCase("engine_2010-06-12.log", "engine_2010-06-12.log.1", "*$D(yyyy-MM-dd).log$J(.)")]


ParseFormatString computes datePos from the original string but edits the escaped one — a literal with regex metachars before $D (e.g. app.$D(yyyy-MM-dd)) corrupts the regex. Not introduced here, but the new trailing-literal support makes such patterns likelier.

Fixed by locating and parsing $D(...) in the escaped format string. Offsets remain correct when literal regex metacharacters appear before the date placeholder.
Added a regression test for app.$D(yyyy-MM-dd).log that verifies date parsing and rebuilding.

Fixed by locating and parsing $D(...) in the escaped format string. Offsets remain correct when literal regex metacharacters appear before the date placeholder. Added a regression test for app.$D(yyyy-MM-dd).log that verifies date parsing and rebuilding. Added cases which cover: - $I with a trailing literal .log, starting from a rotated file. - $D plus $J(.) followed by a trailing .log, also starting from a rotated file.

- Added FlushEscaped() to share duplicated escaping code. - Fixed parsing of $J(<prefix>) from rotated files, so both: - app.log.1 - app.1.log correctly identify index 1. - Made wildcard matching non-greedy and anchored the generated regex to the complete filename. - Updated rebuilding logic to remove an existing prefix plus index together before inserting the new index. - Fixed $D(...) parsing to use positions in the escaped format string, preventing errors when a literal regex character such as . appears before $D.

Pr0metheus2 · 2026-08-29T10:20:41Z

Check please

…mid-filename-index-tags-(app.1.log)

Hirogen · 2026-08-31T14:39:13Z

+        {
+            _ = result.Append(Regex.Escape(segment.ToString()));
+            segment = new StringBuilder();
+        }


extract it to a real function not a function in a function, this only leads to unreadable code

Moved the helper out of EscapeNonvarRegions() into a private class method. (1)

Hirogen · 2026-08-31T14:40:48Z

        StringBuilder result = new();
        StringBuilder segment = new();

+        void FlushEscaped ()


FlushEscaped() covers only case 0, while case 1 and case 3 still inline _ = result.Append(segment); segment = new StringBuilder();.

Added two shared helpers: one appends an escaped literal segment and one appends a placeholder segment. Both also clear the segment, so states 0, 1, and 3 now use the same append & clear behavior. (2)

Hirogen · 2026-08-31T14:42:10Z

        if (_indexGroup != null && _indexGroup.Success)
        {
-            fileName = fileName.Remove(_indexGroup.Index, _indexGroup.Length);
+            var indexPosition = _indexGroup.Index;


Hidden invariant / Feature Envy — indexPosition = _condGroup.Index; indexLength += _condGroup.Length; silently assumes cond is contiguous and immediately precedes index — an invariant set 60 lines away in ParseFormatString. Uncommented; changing either regex breaks the other.

Removed the separate prefix-position calculation. The generated regex now captures the complete index part, including its prefix when present, in one indexPart group. Rebuilding removes and replaces that explicit group directly. (3)

Hirogen · 2026-08-31T14:44:28Z

    }

+    [Test]
+    public void BuildFileName_DatePatternAfterRegexMetacharacter_IncrementsDate ()


BuildFileName_X_Y while the file's existing methods are TestFilenameAnd1/2.

Renamed to TestFilenameDateAfterRegexMetacharacter
(8)

Hirogen · 2026-08-31T14:44:42Z

        _hideZeroIndex = fmt.Contains("$J", StringComparison.Ordinal);
        fmt = fmt.Replace("$I", "(?'index'[\\d]+)", StringComparison.Ordinal);
-        fmt = fmt.Replace("$J", "(?'index'[\\d]*)", StringComparison.Ordinal);
+        var optionalIndexPattern = _condContent != null


optionalIndexPattern reads as "the pattern is optional";

Renamed the pattern variables to indexPattern and conditionalIndexPattern to describe what they match rather than suggesting that the whole pattern is optional. (4)

Hirogen · 2026-08-31T14:45:32Z


        File.Copy(Path.Combine(_testDataDirectory, "app.log"), _logFile);
        File.Copy(Path.Combine(_testDataDirectory, "app.log.1"), _logFile + ".1");
+        File.Copy(Path.Combine(_testDataDirectory, "app.1.log"), Path.Combine(_testDirectory, "app.1.log"));


SetUp copies app.1.log/app.2.log unconditionally, so all four tests now see two extra files though only the new one needs them.

Two copies moved to SingleFileCtor_MultiFileTrue_LoadsIndexBeforeExtension().
(9)

Hirogen · 2026-08-31T14:47:10Z

-        fmt = fmt.Replace("*", ".*", StringComparison.Ordinal);
+        fmt = fmt.Replace("*", ".*?", StringComparison.Ordinal);
        _hideZeroIndex = fmt.Contains("$J", StringComparison.Ordinal);
        fmt = fmt.Replace("$I", "(?'index'[\\d]+)", StringComparison.Ordinal);


"Allow index tags ($J, $I) to appear before a fixed file extension suffix." RolloverFilenameBuilder.cs:189 still maps $I → (?'index'[\d]+); only $J gets the prefix-aware alternation (:190-193). So *$I.log cannot match the active file app.log — SetFileName fails, IsIndexPattern stays false, no chain is built. The one new $I test (RollingNameTest.cs:36, engine1.log→engine2.log) starts from a numbered file and never exercises the failing case.

Updated $I to match an unnumbered active file as well as numbered rotated files.
Added coverage for app.log becoming app1.log after the index is incremented.
(5)

Hirogen · 2026-08-31T14:48:01Z

-$D(&amp;lt;date&amp;gt;) = Datumsmuster
+$D(&lt;date&gt;) = Datumsmuster
 $I = Dateiindexnummer
 $J = Dateiindexnummer, versteckt wenn 09


still reads $J = Dateiindexnummer, versteckt wenn 09 — should be 0. Correct in the English and Chinese files.

Fixed (versteckt wenn null)

Hirogen · 2026-08-31T14:49:43Z

        }

-        fmt = fmt.Replace("*", ".*", StringComparison.Ordinal);
+        fmt = fmt.Replace("*", ".*?", StringComparison.Ordinal);


.→.? and :194 \A…\z anchoring change every mask, not just ones with trailing literals. It holds because RolloverFilenameHandler.cs:39 passes a bare filename, not a path — but that invariant is undocumented and untested; a full path with any non-*-prefixed mask now fails where it previously matched.

Kept the end anchor needed to parse an index before fixed suffix, but removed the start anchor.
This retains compatibility with full paths and non-wildcard masks.
Added a full-path regression test.
(6)

Hirogen · 2026-08-31T14:50:36Z

@@ -116,15 +116,23 @@ public string BuildFileName ()



A saved mask *$J(.).log previously behaved as *$J(.) and matched app.log.1; it now means app.1.log. That's what the spec wants, but existing user settings change

Should the corrected syntax remain this way?

*$J(.) → app.log, app.log.1, app.log.2

*$J(.).log → app.log, app.1.log, app.2.log

If so, should I migrate existing saved *$J(.).log masks to *$J(.)?
(7)

Added settings migration for corrected MultiFile mask behavior.

To preserve compatibility, settings version 2 migrates existing saved *$J(.).log preferences to *$J(.). New version-2 settings keep both patterns distinct:

*$J(.) for app.log.1

*$J(.).log for app.1.log

Added tests for migrating old settings and for ensuring new settings are not changed.

Added a settings migration for the corrected MultiFile mask behavior. Older LogExpert versions accidentally ignored the trailing .log in saved *$J(.).log masks, so users effectively got app.log.1 style rollover files. The parser now correctly supports app.1.log for that mask, but this would otherwise change existing users’ saved behavior. To preserve compatibility, settings version 2 migrates existing saved *$J(.).log preferences to *$J(.). New version-2 settings keep both patterns distinct: - *$J(.) for app.log.1 - *$J(.).log for app.1.log Added tests for migrating old settings and for ensuring new settings are not changed.

Hirogen · 2026-09-05T10:59:52Z

+    private static void MigrateToV2 (Settings settings)
+    {
+        // Before the parser fix, the trailing .log in this pattern was ignored, making it behave
+        // like *$J(.). Preserve that behavior for existing saved preferences.


rewrites a saved *$J(.).log to *$J(.). A user who already saved that pattern hoping for the new grouping, which is exactly the situation, gets silently reverted to the old behaviour. It bumps the schema version for every user and does not touch per-file persisted patterns, so the "preserve old behaviour" rationale is inconsistent.

(1) OK, I've removed the settings migration completely:

Schema version restored to 1.

Removed MigrateToV2().

Removed its tests and test-only import.

Existing saved *$J(.).log masks now retain the corrected new meaning.

Hirogen · 2026-09-05T11:00:59Z

+        fmt = fmt.Replace("$I", indexPattern, StringComparison.Ordinal);
+        fmt = fmt.Replace("$J", conditionalIndexPattern, StringComparison.Ordinal);
+
+        _regex = new Regex(fmt + @"\z");


$I digit class went from one-or-more to zero-or-more. A plain app.log now matches *$I.log as index 0, so $I behaves like $J when the opened file has no digits.

*$J(.) no longer behaves exactly as before. The lazy wildcard plus the end anchor mean opening engine.log.1 now yields index 1. Previously it yielded index 0 and generated engine.log.1.1. This is a fix, but it is an unstated behaviour change.

The FlushEscaped method named in the PR description does not exist. The helpers are AppendEscapedSegment and AppendSegment.

(2) Restored $I behavior:

$I again requires one or more digits ([\d]+).

app.log no longer matches *$I.log.

$J retains its zero-index handling and the rotated-file parsing fix.

The corrected behavior for opening engine.log.1 is kept; describe that explicitly in the PR rather than treating it as invisible behavior.

Note: I cannot find FlushEscaped in the current PR description.

Hirogen · 2026-09-05T11:02:50Z

+    [Test]
+    public void SingleFileCtor_MultiFileTrue_LoadsIndexBeforeExtension ()
+    {
+        var options = new MultiFileOptions { FormatPattern = "*$J(.).log" };


The $I case is only tested from the active file, despite the PR description claiming a rotated-file $I test.

(3) Updated the $I test to start from a rotated filename:
engine1.log → engine2.log
with *$I.log.

Hirogen · 2026-09-05T11:03:52Z

    {
        return new LogfileReader(
            _logFile,
            new EncodingOptions { Encoding = Encoding.UTF8 },


opens the active file only. Opening a rotated file directly never picks up lower siblings because the name-list walk only goes upward. That is pre-existing, but the spec is only met when the active file is opened.

No round-trip test for *$J(.).log.

My understanding is that this is intended behavior: when a rotated file such as app.1.log is opened, MultiFile loads that file and higher indexes only. To load the complete chain including lower indexes, the user should open the active file (app.log). If user want to open log file from specified date only and further, i.e. app.8.log, this is a useful behaviour and in this case app will load app.8.log -> app.9.log etc.

Please confirm whether this is correct. If MultiFile should also discover lower-index siblings when a rotated file is opened directly, I can extend the handler and add coverage for that behavior.

Pr0metheus2 added 2 commits August 26, 2026 12:07

test: add multi-file index test data

3989ae2

Hirogen requested changes Aug 28, 2026

View reviewed changes

Pr0metheus2 added 4 commits August 29, 2026 11:36

Fixed from when to wenn

15c3d96

Removed old commented lines

905c979

Merge branch 'LogExperts:Development' into fix/MultiFile-support-for-…

2a5e1ed

…mid-filename-index-tags-(app.1.log)

Hirogen requested changes Aug 31, 2026

View reviewed changes

Pr0metheus2 added 4 commits September 1, 2026 15:35

Update Resources.de.resx

ae20df6

Update LogfileReaderMultiFileFlagTests.cs

00d4c26

Update

d23d228

Hirogen requested changes Sep 5, 2026

View reviewed changes

Change update

a1320e7

Uh oh!

Conversation

Pr0metheus2 commented Aug 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Pr0metheus2 commented Aug 29, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Pr0metheus2 Sep 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Pr0metheus2 commented Aug 26, 2026 •

edited

Loading

Pr0metheus2 Sep 1, 2026 •

edited

Loading

Pr0metheus2 Sep 8, 2026 •

edited

Loading