Compare commits

...
3 Commits
Author SHA1 Message Date
Youmu Liu 57587bec16 use new exit lable 2026-09-14 18:21:23 +08:00
Youmu Liu 9e9078ced1 meet difficult problem on COUNT node 2026-09-11 17:30:31 +08:00
Youmu Liu 5a3286dc21 meet difficult problem on COUNT node 2026-09-11 17:29:42 +08:00
8 changed files with 213 additions and 123 deletions
@@ -16,12 +16,15 @@ namespace Nitload.RegularAnalysis.Text
internal partial class TextRegexSourceGen
{
internal enum ExitMode
internal enum ExitKind
{
Break,
Continue,
Goto,
MatchFailed,
MatchEnd,
CountRetry,
CountResume,
AltNext,
AltMatched,
}
@@ -29,25 +32,28 @@ namespace Nitload.RegularAnalysis.Text
{
public ExitLabel(
ExitMode mode,
ExitKind kind,
int order
)
{
Mode = mode;
Kind = kind;
LabelOrder = order;
}
public ExitMode Mode { get; init; }
public ExitKind Kind { get; init; }
public int LabelOrder { get; init; }
public string GetLabelName()
{
return Mode switch
return Kind switch
{
ExitMode.Break => $@"__break_{LabelOrder}",
ExitMode.Continue => $@"__continue_{LabelOrder}",
ExitMode.Goto => $@"__goto_{LabelOrder}",
ExitKind.MatchFailed => $@"__match_failed_{LabelOrder}",
ExitKind.MatchEnd => $@"__match_end_{LabelOrder}",
ExitKind.CountRetry => $@"__count_retry_{LabelOrder}",
ExitKind.CountResume => $@"__count_resume_{LabelOrder}",
ExitKind.AltNext => $@"__alt_next_{LabelOrder}",
ExitKind.AltMatched => $@"__alt_matched_{LabelOrder}",
_ => throw new NotImplementedException(),
};
}
@@ -14,20 +14,31 @@ namespace Nitload.RegularAnalysis.Text
internal partial class TextRegexSourceGen
{
/// <summary>
/// Emits a subtree and updates the retry target for the rest of the current path.
/// </summary>
/// <remarks>
/// local_fail is the owning region's fallback. backtrack starts at that fallback and
/// advances when a COUNT registers a retry point. Returning from recursion does not
/// discard it. Mutually exclusive branches must use separate backtrack variables.
/// </remarks>
internal static void GenerateMethodSource(
RegexAstNode ast,
in SequenceEmitter<string> sc,
in List<SequenceEmitter<string>> func_list,
in SourceGenContext context,
in ExitLabel exit_label = default,
in ExitLabel local_fail,
ref ExitLabel backtrack,
bool is_find = false
)
{
// local_fail belongs to this region. COUNT updates only the current path's
// backtrack target; CONCAT passes that update on to every later sibling.
RegexAstNode root = ast;
if (root.Label == RegexElementLabel.EPSILON) {
return;
}
if (TryGenerateSpecialSource(root, sc, func_list, context, exit_label)) {
if (TryGenerateSpecialSource(root, sc, func_list, context, local_fail, ref backtrack)) {
return;
}
RegexElementLabel op = root.Label & (RegexElementLabel.VAST_OPERATOR);
@@ -36,16 +47,16 @@ namespace Nitload.RegularAnalysis.Text
switch (op)
{
case RegexElementLabel.CONCAT:
GenerateConcat(root, sc, func_list, context, exit_label);
GenerateConcat(root, sc, func_list, context, local_fail, ref backtrack);
break;
case RegexElementLabel.COUNT:
GenerateCount(root, sc, func_list, context, exit_label);
GenerateCount(root, sc, func_list, context, local_fail, ref backtrack);
break;
case RegexElementLabel.GROUP_BEGIN:
GenerateGroup(root, sc, func_list, context, exit_label);
GenerateGroup(root, sc, func_list, context, local_fail, ref backtrack);
break;
case RegexElementLabel.ALT:
GenerateAlt(root, sc, func_list, context, exit_label);
GenerateAlt(root, sc, func_list, context, backtrack);
break;
default:
@@ -59,13 +70,13 @@ namespace Nitload.RegularAnalysis.Text
switch (op)
{
case RegexElementLabel.UNIT:
GenerateUnit(root, sc, context, exit_label);
GenerateUnit(root, sc, context, backtrack);
break;
case RegexElementLabel.UNIT_SET:
GenerateUnitSet(root, sc, context, exit_label);
GenerateUnitSet(root, sc, context, backtrack);
break;
case RegexElementLabel.UNIT_SPAN:
GenerateUnitSpan(root, sc, context, exit_label);
GenerateUnitSpan(root, sc, context, backtrack);
break;
default:
break;
@@ -221,7 +232,8 @@ namespace Nitload.RegularAnalysis.Text
bool is_greedy,
[NotNull] in SequenceEmitter<string> sc,
in SourceGenContext context,
in ExitLabel exit
in ExitLabel exit,
ref ExitLabel backtrack
)
{
if (min < 0 || (max >= 0 && max < min)) {
@@ -300,23 +312,54 @@ namespace Nitload.RegularAnalysis.Text
}
int additional_limit = max < 0 ? -1 : max - min;
if (is_greedy && additional_limit != 0)
{
_ = additional_limit < 0 ?
sc.AppendLine($"while (!{remain_span}.IsEmpty && {remain_span}[0] == {expected_literal})")
.AppendLine("{")
.AppendLine($"{remain_span} = {remain_span}.Slice(1);")
.AppendLine("}") :
sc.AppendLine($"for (int {additional_count} = 0; {additional_count} < {additional_limit} && !{remain_span}.IsEmpty && {remain_span}[0] == {expected_literal}; {additional_count}++)")
.AppendLine("{")
.AppendLine($"{remain_span} = {remain_span}.Slice(1);")
.AppendLine("}");
} else if (!is_greedy && additional_limit != 0)
{
// Lazy COUNT deliberately stops at Min. Its optional tail is consumed only when
// a future continuation attempt fails and requests another repetition.
_ = sc.AppendLine("// TODO(NGRA): Retry the lazy COUNT tail from its continuation.");
if (additional_limit == 0) {
return;
}
// Save the optional portion once. Its counter is shared by scanning and retries.
string additional_entry = $"__additional_entry_{scratch_scope_index}";
ExitLabel retry = context.RequestExitLabel(ExitKind.CountRetry);
ExitLabel resume = context.RequestExitLabel(ExitKind.CountResume);
_ = sc.AppendLine($"{ROS_char} {additional_entry} = {remain_span};")
.AppendLine($"int {additional_count} = 0;");
if (is_greedy)
{
string limit = additional_limit < 0 ? string.Empty :
$"{additional_count} < {additional_limit} && ";
_ = sc.AppendLine($"while ({limit}!{remain_span}.IsEmpty && {remain_span}[0] == {expected_literal})")
.AppendLine("{")
.AppendLine($"{remain_span} = {remain_span}.Slice(1);")
.AppendLine($"{additional_count}++;")
.AppendLine("}");
}
// Initial success skips retry handling. Later failures jump back here from anywhere
// in this path; exhaustion forwards to the target inherited before this COUNT.
_ = sc.AppendLine($"goto {resume.GetLabelName()};")
.AppendLine($"{retry.GetLabelName()}: {{}}");
if (is_greedy)
{
_ = sc.AppendLine($"if ({additional_count} == 0)")
.AppendLine("{")
.AppendLine($"{remain_span} = {additional_entry};")
.AppendLine($"goto {exit_label_node};")
.AppendLine("}")
.AppendLine($"{additional_count}--;");
} else
{
string limit = additional_limit < 0 ? string.Empty :
$"{additional_count} == {additional_limit} || ";
_ = sc.AppendLine($"{remain_span} = {additional_entry}.Slice({additional_count});")
.AppendLine($"if ({limit}{remain_span}.IsEmpty || {remain_span}[0] != {expected_literal})")
.AppendLine("{")
.AppendLine($"{remain_span} = {additional_entry};")
.AppendLine($"goto {exit_label_node};")
.AppendLine("}")
.AppendLine($"{additional_count}++;");
}
_ = sc.AppendLine($"{remain_span} = {additional_entry}.Slice({additional_count});")
.AppendLine($"{resume.GetLabelName()}: {{}}");
backtrack = retry;
}
private static void GenerateUnitSetCount(
@@ -400,7 +443,8 @@ namespace Nitload.RegularAnalysis.Text
bool is_greedy,
[NotNull] in SequenceEmitter<string> sc,
in SourceGenContext context,
in ExitLabel exit
in ExitLabel exit,
ref ExitLabel backtrack
)
{
if (min < 0 || (max >= 0 && max < min)) {
@@ -510,27 +554,54 @@ namespace Nitload.RegularAnalysis.Text
}
int additional_limit = max < 0 ? -1 : max - min;
if (is_greedy && additional_limit != 0)
{
if (additional_limit < 0)
{
_ = sc.AppendLine($"while ({remain_span}.{starts_with}({expected_literal}))")
.AppendLine("{")
.AppendLine($"{remain_span} = {remain_span}.Slice({literal.Length});")
.AppendLine("}");
} else
{
_ = sc.AppendLine($"for (int {additional_count} = 0; {additional_count} < {additional_limit} && {remain_span}.{starts_with}({expected_literal}); {additional_count}++)")
.AppendLine("{")
.AppendLine($"{remain_span} = {remain_span}.Slice({literal.Length});")
.AppendLine("}");
}
} else if (!is_greedy && additional_limit != 0)
{
// Lazy COUNT deliberately stops at Min. Its optional tail is consumed only when
// a future continuation attempt fails and requests another repetition.
_ = sc.AppendLine("// TODO(NGRA): Retry the lazy COUNT tail from its continuation.");
if (additional_limit == 0) {
return;
}
// Save the optional portion once. Its counter is shared by scanning and retries.
string additional_entry = $"__additional_entry_{scratch_scope_index}";
ExitLabel retry = context.RequestExitLabel(ExitKind.CountRetry);
ExitLabel resume = context.RequestExitLabel(ExitKind.CountResume);
_ = sc.AppendLine($"{ROS_char} {additional_entry} = {remain_span};")
.AppendLine($"int {additional_count} = 0;");
if (is_greedy)
{
string limit = additional_limit < 0 ? string.Empty :
$"{additional_count} < {additional_limit} && ";
_ = sc.AppendLine($"while ({limit}{remain_span}.{starts_with}({expected_literal}))")
.AppendLine("{")
.AppendLine($"{remain_span} = {remain_span}.Slice({literal.Length});")
.AppendLine($"{additional_count}++;")
.AppendLine("}");
}
// Initial success skips retry handling. Later failures jump back here from anywhere
// in this path; exhaustion forwards to the target inherited before this COUNT.
_ = sc.AppendLine($"goto {resume.GetLabelName()};")
.AppendLine($"{retry.GetLabelName()}: {{}}");
if (is_greedy)
{
_ = sc.AppendLine($"if ({additional_count} == 0)")
.AppendLine("{")
.AppendLine($"{remain_span} = {additional_entry};")
.AppendLine($"goto {exit_label_node};")
.AppendLine("}")
.AppendLine($"{additional_count}--;");
} else
{
string limit = additional_limit < 0 ? string.Empty :
$"{additional_count} == {additional_limit} || ";
_ = sc.AppendLine($"{remain_span} = {additional_entry}.Slice({additional_count} * {literal.Length});")
.AppendLine($"if ({limit}!{remain_span}.{starts_with}({expected_literal}))")
.AppendLine("{")
.AppendLine($"{remain_span} = {additional_entry};")
.AppendLine($"goto {exit_label_node};")
.AppendLine("}")
.AppendLine($"{additional_count}++;");
}
_ = sc.AppendLine($"{remain_span} = {additional_entry}.Slice({additional_count} * {literal.Length});")
.AppendLine($"{resume.GetLabelName()}: {{}}");
backtrack = retry;
}
#endregion
@@ -552,7 +623,7 @@ namespace Nitload.RegularAnalysis.Text
in ExitLabel exit_label
)
{
ExitLabel success = context.RequestExitLabel(ExitMode.Goto);
ExitLabel success = context.RequestExitLabel(ExitKind.AltMatched);
RegexAstNode current = node.Right;
while (!RegexAstNode.IsNullOrEmpty(current))
{
@@ -561,7 +632,7 @@ namespace Nitload.RegularAnalysis.Text
}
_ = sc.AppendLine($"goto {exit_label.GetLabelName()};")
.AppendLine($"{success.GetLabelName()}: ;");
.AppendLine($"{success.GetLabelName()}: {{}}");
}
private static void GenerateGroup(
@@ -569,7 +640,8 @@ namespace Nitload.RegularAnalysis.Text
[NotNull] in SequenceEmitter<string> sc,
in List<SequenceEmitter<string>> func_list,
in SourceGenContext context,
in ExitLabel exit_label
in ExitLabel local_fail,
ref ExitLabel backtrack
)
{
RegexAstNode left = node.Left;
@@ -583,10 +655,10 @@ namespace Nitload.RegularAnalysis.Text
// TODO(NGRA): Snapshot the group start cursor and restore it when the child fails.
if (!RegexAstNode.IsNullOrEmpty(left))
{
GenerateMethodSource(left, sc, func_list, context, exit_label);
GenerateMethodSource(left, sc, func_list, context, local_fail, ref backtrack);
} else
if (!RegexAstNode.IsNullOrEmpty(right)) {
GenerateMethodSource(right, sc, func_list, context, exit_label);
GenerateMethodSource(right, sc, func_list, context, local_fail, ref backtrack);
}
string group_name = config.Name;
@@ -601,14 +673,15 @@ namespace Nitload.RegularAnalysis.Text
[NotNull] in SequenceEmitter<string> sc,
in List<SequenceEmitter<string>> func_list,
in SourceGenContext context,
in ExitLabel exit_label = default
in ExitLabel local_fail,
ref ExitLabel backtrack
)
{
if (!RegexAstNode.IsNullOrEmpty(node.Left)) {
GenerateMethodSource(node.Left, sc, func_list, context, exit_label);
GenerateMethodSource(node.Left, sc, func_list, context, local_fail, ref backtrack);
}
if (!RegexAstNode.IsNullOrEmpty(node.Right)) {
GenerateMethodSource(node.Right, sc, func_list, context, exit_label);
GenerateMethodSource(node.Right, sc, func_list, context, local_fail, ref backtrack);
}
}
@@ -617,17 +690,29 @@ namespace Nitload.RegularAnalysis.Text
[NotNull] in SequenceEmitter<string> sc,
in List<SequenceEmitter<string>> func_list,
in SourceGenContext context,
in ExitLabel exit_label
in ExitLabel local_fail,
ref ExitLabel backtrack
)
{
RegexAstNode value = RegexAstNode.IsNullOrEmpty(node.Left) ? node.Right : node.Left;
if (RegexAstNode.IsNullOrEmpty(value)) {
throw new InvalidOperationException("COUNT requires an operand.");
}
// Inspect transparent CONCAT wrappers without rewriting the repetition body.
while (value.Label == RegexElementLabel.CONCAT)
{
bool left_empty = RegexAstNode.IsNullOrEmpty(value.Left);
bool right_empty = RegexAstNode.IsNullOrEmpty(value.Right);
if (left_empty == right_empty) {
break;
}
value = left_empty ? value.Right : value.Left;
}
RegexCountPayload payload = node.PayloadAs<RegexCountPayload>();
int min = payload.Min;
int max = payload.Max;
bool is_greedy = payload.IsGreedy;
ExitLabel failure = backtrack;
// TODO(NGRA-NAMING): Reserve exclusive names for future repetition/rollback state
// across complex COUNT retries; recursive children must use the same method pool.
@@ -635,20 +720,20 @@ namespace Nitload.RegularAnalysis.Text
{
// TODO(NGRA): Emit min/max repetition, zero-width progress protection, greedy vs.
// lazy ordering, and per-attempt rollback for a complex COUNT body.
GenerateMethodSource(value, sc, func_list, context, exit_label);
GenerateMethodSource(value, sc, func_list, context, local_fail, ref backtrack);
} else
{
switch (value.Label & RegexElementLabel.SEQUENCE)
{
case RegexElementLabel.UNIT:
GenerateUnitCount(value, min, max, is_greedy, sc, context, exit_label);
break;
GenerateUnitCount(value, min, max, is_greedy, sc, context, failure, ref backtrack);
return;
case RegexElementLabel.UNIT_SET:
GenerateUnitSetCount(value, min, max, is_greedy, sc, context, exit_label);
GenerateUnitSetCount(value, min, max, is_greedy, sc, context, failure);
break;
case RegexElementLabel.UNIT_SPAN:
GenerateUnitSpanCount(value, min, max, is_greedy, sc, context, exit_label);
break;
GenerateUnitSpanCount(value, min, max, is_greedy, sc, context, failure, ref backtrack);
return;
default:
// TODO(NGRA): Diagnose malformed COUNT children instead of emitting an
@@ -656,7 +741,6 @@ namespace Nitload.RegularAnalysis.Text
break;
}
}
}
#endregion
@@ -36,7 +36,7 @@ namespace Nitload.RegularAnalysis.Text
public const string _remain_span = "__remain";
private int _max_exit_lable;
private int _exit_label_count;
private int _scratch_scope_count;
@@ -54,12 +54,12 @@ namespace Nitload.RegularAnalysis.Text
}
public ExitLabel RequestExitLabel(
ExitMode exitMode
ExitKind kind
)
{
int index = _max_exit_lable;
_max_exit_lable++;
return new(exitMode, index);
int index = _exit_label_count;
_exit_label_count++;
return new(kind, index);
}
public int RequestScratchScopeIndex()
@@ -17,7 +17,8 @@ namespace Nitload.RegularAnalysis.Text
in SequenceEmitter<string> sc,
in List<SequenceEmitter<string>> func_list,
in SourceGenContext context,
in ExitLabel exit_label
in ExitLabel local_fail,
ref ExitLabel backtrack
)
{
// A special rule owns the entire matched subtree. Returning false leaves emission
@@ -25,7 +26,7 @@ namespace Nitload.RegularAnalysis.Text
return node.Label switch
{
RegexElementLabel.CONCAT => TryGenerateFindConcat(node, sc, func_list, context,
exit_label),
local_fail, ref backtrack),
_ => false,
};
}
@@ -166,10 +167,12 @@ namespace Nitload.RegularAnalysis.Text
in ExitLabel success
)
{
ExitLabel failure = context.RequestExitLabel(ExitMode.Goto);
ExitLabel failure = context.RequestExitLabel(ExitKind.AltNext);
string entry = $"__alt_entry_{context.RequestScratchScopeIndex()}";
_ = sc.AppendLine($"{ROS_char} {entry} = {context.RemainSpan};");
GenerateMethodSource(node, sc, func_list, context, failure);
// Keep branch-local retries isolated from both siblings and the caller's path.
ExitLabel branch_backtrack = failure;
GenerateMethodSource(node, sc, func_list, context, failure, ref branch_backtrack);
_ = sc.AppendLine($"goto {success.GetLabelName()};")
.AppendLine($"{failure.GetLabelName()}:")
.AppendLine($"{context.RemainSpan} = {entry};");
@@ -187,7 +190,8 @@ namespace Nitload.RegularAnalysis.Text
in SequenceEmitter<string> sc,
in List<SequenceEmitter<string>> func_list,
in SourceGenContext context,
in ExitLabel exit_label
in ExitLabel local_fail,
ref ExitLabel backtrack
)
{
RegexAstNode left = node.Left;
@@ -202,7 +206,7 @@ namespace Nitload.RegularAnalysis.Text
{
// TODO(NGRA): Emit the unnamed FIND candidate loop and retry the right subtree
// until it succeeds or no anchor candidate remains.
GenerateMethodSource(right, sc, func_list, context, exit_label, true);
GenerateMethodSource(right, sc, func_list, context, local_fail, ref backtrack, true);
return true;
}
@@ -212,7 +216,7 @@ namespace Nitload.RegularAnalysis.Text
{
// TODO(NGRA): Emit named FIND retry and capture the text preceding the
// successful anchor candidate.
GenerateMethodSource(right, sc, func_list, context, exit_label, true);
GenerateMethodSource(right, sc, func_list, context, local_fail, ref backtrack, true);
} else
{
// TODO(NGRA): Capture the remaining input for a terminal named dot-star.
@@ -172,7 +172,15 @@ namespace Nitload.RegularAnalysis.Text
SequenceEmitter<string> methodSource = new();
List<SequenceEmitter<string>> localFunctions = [];
GenerateMethodSource(root, methodSource, localFunctions, generationContext);
ExitLabel local_fail = generationContext.RequestExitLabel(ExitKind.MatchFailed);
ExitLabel backtrack = local_fail;
ExitLabel end = generationContext.RequestExitLabel(ExitKind.MatchEnd);
GenerateMethodSource(root, methodSource, localFunctions, generationContext,
local_fail, ref backtrack);
_ = methodSource.AppendLine($"goto {end.GetLabelName()};")
.AppendLine($"{local_fail.GetLabelName()}: {{}}")
.AppendLine("// TODO(NGRA): Produce the root failure result here.")
.AppendLine($"{end.GetLabelName()}: {{}}");
foreach (SequenceEmitter<string> localFunction in localFunctions) {
_ = methodSource.AppendLine()
.AppendLine()
+3 -12
View File
@@ -24,17 +24,8 @@ namespace Nitload
private static string testExpPath = @"D:\Programmer\stepcode\data\ap203\ap203.exp";
internal static unsafe void Main(
string[] args
)
internal static unsafe void Main()
{
// Select the managed-only emitter preview before initializing the normal runtime.
if (args.Length > 0 && args[0] == "--regex-preview")
{
Environment.ExitCode = TestRegexGenerate.Run(args[1..]);
return;
}
try
{
World.Initialize(new()
@@ -46,8 +37,8 @@ namespace Nitload
// .UseModule<GraphicModule>()
.InitializeModule();
// TestRegexAnalyse.Run();
_ = TestRegexGenerate.Run("abc|123");
TestRegexAnalyse.Run();
_ = TestRegexGenerate.Run("abc1{1,2}23");
// TestHlsf();
+1 -1
View File
@@ -52,7 +52,7 @@ namespace Nitload
}
}
[GeneratedRegex(@"123.*\w123", RegexOptions.ExplicitCapture)]
[GeneratedRegex(@"(a{2,4}bc)d", RegexOptions.ExplicitCapture)]
internal static partial Regex GetRegex();
}
+20 -23
View File
@@ -1,6 +1,6 @@
// -----------------------------------------------------------------------------
// file="TestRegexGenerate"
// Project: Nitload
// Project: StgSharp
// Copyright (c) Nitload.
// SPDX-License-Identifier: MIT
// -----------------------------------------------------------------------------
@@ -16,29 +16,19 @@ namespace Nitload
{
internal static class TestRegexGenerate
{
public static int Run(params string[] args)
{
if (args.Length > 1) {
throw new ArgumentException("Expected at most one regex pattern.", nameof(args));
}
if (args.Length == 1)
{
Print(args[0]);
return 0;
}
while (true)
{
Console.Write("regex> ");
string? pattern = Console.ReadLine();
if (pattern is null or ":exit") {
return 0;
}
Print(pattern);
}
public static int Run(
string arg
)
{
Console.WriteLine($"Text regex to be tested: {arg} :\n\n\n");
Print(arg);
return 0;
}
private static void Print(string pattern)
private static void Print(
string pattern
)
{
TextRegexSource analysis = RegexAnalyzer.Analyze(pattern);
if (analysis.AnalyseException is not null) {
@@ -48,8 +38,14 @@ namespace Nitload
SequenceEmitter<string> body = new();
List<SequenceEmitter<string>> functions = [];
SourceGenContext context = new();
ExitLabel exit = context.RequestExitLabel(ExitMode.Goto);
GenerateMethodSource(analysis.Ast.Root, body, functions, context, exit);
ExitLabel local_fail = context.RequestExitLabel(ExitKind.MatchFailed);
ExitLabel backtrack = local_fail;
ExitLabel end = context.RequestExitLabel(ExitKind.MatchEnd);
GenerateMethodSource(analysis.Ast.Root, body, functions, context, local_fail, ref backtrack);
_ = body.AppendLine($"goto {end.GetLabelName()};")
.AppendLine($"{local_fail.GetLabelName()}: {{}}")
.AppendLine("// TODO(NGRA): Produce the root failure result here.")
.AppendLine($"{end.GetLabelName()}: {{}}");
foreach (string line in body.ToList()) {
Console.WriteLine(line);
@@ -61,5 +57,6 @@ namespace Nitload
}
}
}
}
}