/// <summary>
/// Incremental parsing of a set of processed tokens lines changes
/// </summary>
internal static IList <DocumentChange <ICodeElementsLine> > ParseProcessedTokensLinesChanges(TextSourceInfo textSourceInfo, ISearchableReadOnlyList <CodeElementsLine> documentLines, IList <DocumentChange <IProcessedTokensLine> > processedTokensLinesChanges, PrepareDocumentLineForUpdate prepareDocumentLineForUpdate, TypeCobolOptions compilerOptions, PerfStatsForParserInvocation perfStatsForParserInvocation)
{
// Collect all changes applied to the processed tokens lines during the incremental scan
IList <DocumentChange <ICodeElementsLine> > codeElementsLinesChanges = new List <DocumentChange <ICodeElementsLine> >();
// There are 2 reasons to re-parse a tokens line after a change :
// 1. The tokens line changed : these lines were already reset during the previous steps
// 2. If a tokens line that changed was involved in the parsing of a multiline code element, the whole group of lines must be parsed again
// --- PREPARATION PHASE : identify all parse sections where code elements need to be refreshed ---
IList <ParseSection> refreshParseSections = null;
ParseSection largestRefreshParseSection = null;
// Iterate over all processed tokens changes detected by the PreprocessorStep :
// - refresh all the adjacent lines participating in a CodeElement
// - register the start and stop token for all sections of the document which need to be parsed again
if (processedTokensLinesChanges != null && processedTokensLinesChanges.Count > 0)
{
// If the document was cleared, everything must be parsed again
if (processedTokensLinesChanges[0].Type != DocumentChangeType.DocumentCleared)
{
refreshParseSections = new List <ParseSection>();
ParseSection lastParseSection = null;
foreach (DocumentChange <IProcessedTokensLine> tokensChange in processedTokensLinesChanges)
{
if (lastParseSection == null || tokensChange.LineIndex > lastParseSection.StopLineIndex)
{
lastParseSection = CheckIfAdjacentLinesNeedRefresh(tokensChange.Type, tokensChange.LineIndex, documentLines, prepareDocumentLineForUpdate, codeElementsLinesChanges, lastParseSection);
refreshParseSections.Add(lastParseSection);
}
}
}
}
if (refreshParseSections != null)
{
//After getting all the parts refreshed, get the largest part that has been refreshed
var minParseSection = refreshParseSections.OrderBy(p => p.StartLineIndex).First();
var maxParseSection = refreshParseSections.OrderByDescending(p => p.StopLineIndex).First();
largestRefreshParseSection = new ParseSection(minParseSection.StartLineIndex,
minParseSection.StartToken, maxParseSection.StopLineIndex, maxParseSection.StopToken,
maxParseSection.StopTokenIsFirstTokenOfTheLine);
}
// --- INITIALIZE ANTLR CodeElements parser ---
// Create a token iterator on top of pre-processed tokens lines
ITokensLinesIterator tokensIterator = ProcessedTokensDocument.GetProcessedTokensIterator(textSourceInfo, documentLines);
// Create an Antlr compatible token source on top of the token iterator
TokensLinesTokenSource tokenSource = new TokensLinesTokenSource(
textSourceInfo.Name,
tokensIterator);
// Init parser
TokensLinesTokenStream tokenStream = new TokensLinesTokenStream(tokenSource, Token.CHANNEL_SourceTokens);
CodeElementsParser cobolParser = new CodeElementsParser(tokenStream);
// REVERT TO STD PARSER ==> TracingCobolParser cobolParser = new TracingCobolParser(tokenStream);
// Optionnaly activate Antlr Parser performance profiling
// WARNING : use this in a single-treaded context only (uses static field)
if (AntlrPerformanceProfiler == null && perfStatsForParserInvocation.ActivateDetailedAntlrPofiling)
{
AntlrPerformanceProfiler = new AntlrPerformanceProfiler(cobolParser);
}
if (AntlrPerformanceProfiler != null)
{
// Replace the generated parser by a subclass which traces all rules invocations
cobolParser = new CodeElementsTracingParser(tokenStream);
var tokensCountIterator = ProcessedTokensDocument.GetProcessedTokensIterator(textSourceInfo, documentLines);
AntlrPerformanceProfiler.BeginParsingFile(textSourceInfo, tokensCountIterator);
}
// Customize error recovery strategy
IAntlrErrorStrategy cobolErrorStrategy = new CodeElementErrorStrategy();
cobolParser.ErrorHandler = cobolErrorStrategy;
// Register all parse errors in a list in memory
ParserDiagnosticErrorListener errorListener = new ParserDiagnosticErrorListener();
cobolParser.RemoveErrorListeners();
cobolParser.AddErrorListener(errorListener);
// Prepare to analyze the parse tree
ParseTreeWalker walker = new ParseTreeWalker();
CodeElementBuilder codeElementBuilder = new CodeElementBuilder();
codeElementBuilder.Dispatcher = new CodeElementDispatcher();
codeElementBuilder.Dispatcher.CreateListeners();
// --- INCREMENTAL PARSING ---
// In case of incremental parsing, parse only the code sections we need to refresh
if (largestRefreshParseSection != null)
{
// Seek just before the next code element starting token
tokenStream.SeekToToken(largestRefreshParseSection.StartToken);
tokenStream.StartLookingForStopToken(largestRefreshParseSection.StopToken);
//Remove all the code elements for the future line to parse.
for (int i = largestRefreshParseSection.StartLineIndex;
i < (largestRefreshParseSection.StopLineIndex == documentLines.Count - 1 && largestRefreshParseSection.StopToken == null //If the last index is equals to number of line in document, make sure to also reset the last line, otherwise, reset lines normally.
? largestRefreshParseSection.StopLineIndex + 1
: largestRefreshParseSection.StopLineIndex);
i++)
{
if (documentLines[i].CodeElements != null)
{
documentLines[i].ResetCodeElements();
}
}
}
// Reset parsing error diagnostics
cobolErrorStrategy.Reset(cobolParser);
// Try to parse code elements :
// - starting with the current parse section Start token
// - ending with the current parse section Stop token
CodeElementsParser.CobolCodeElementsContext codeElementsParseTree = null;
try
{
perfStatsForParserInvocation.OnStartAntlrParsing();
if (AntlrPerformanceProfiler != null)
{
AntlrPerformanceProfiler.BeginParsingSection();
}
codeElementsParseTree = cobolParser.cobolCodeElements();
if (AntlrPerformanceProfiler != null)
{
AntlrPerformanceProfiler.EndParsingSection(codeElementsParseTree.ChildCount);
}
perfStatsForParserInvocation.OnStopAntlrParsing(
AntlrPerformanceProfiler != null
? (int)AntlrPerformanceProfiler.CurrentFileInfo.DecisionTimeMs
: 0,
AntlrPerformanceProfiler != null
? AntlrPerformanceProfiler.CurrentFileInfo.RuleInvocations.Sum()
: 0);
}
catch (Exception e)
{
var currentToken = (Token)cobolParser.CurrentToken;
CodeElementsLine codeElementsLine = GetCodeElementsLineForToken(currentToken);
if (codeElementsLine != null)
{
codeElementsLine.AddParserDiagnostic(new TokenDiagnostic(MessageCode.ImplementationError,
currentToken, currentToken.Line, e));
}
}
if (codeElementsParseTree != null)
{
// If the parse tree is not empty
if (codeElementsParseTree.codeElement() != null && codeElementsParseTree.codeElement().Length > 0)
{
// Analyze the parse tree for each code element
foreach (var codeElementParseTree in codeElementsParseTree.codeElement())
{
// Get the first line that was parsed
var tokenStart = (Token)codeElementParseTree.Start;
CodeElementsLine codeElementsLine = GetCodeElementsLineForToken(tokenStart);
if (codeElementsLine == null)
{
continue;
}
// Register that this line was updated
// COMMENTED FOR THE SAKE OF PERFORMANCE -- SEE ISSUE #160
//int updatedLineIndex = documentLines.IndexOf(codeElementsLine, codeElementsLine.LineIndex);
//codeElementsLinesChanges.Add(new DocumentChange<ICodeElementsLine>(DocumentChangeType.LineUpdated, updatedLineIndex, codeElementsLine));
codeElementsLinesChanges.Add(
new DocumentChange <ICodeElementsLine>(DocumentChangeType.LineUpdated,
codeElementsLine.LineIndex, codeElementsLine));
perfStatsForParserInvocation.OnStartTreeBuilding();
// Visit the parse tree to build a first class object representing the code elements
try
{
walker.Walk(codeElementBuilder, codeElementParseTree);
}
catch (Exception ex)
{
var code = MessageCode.ImplementationError;
int line = 0;
int start = 0;
int stop = 0;
if (codeElementsLine.SourceTokens != null && codeElementsLine.SourceTokens.Count > 0)
{
start = codeElementsLine.SourceTokens[0].StartIndex;
stop =
codeElementsLine.SourceTokens[codeElementsLine.SourceTokens.Count - 1].StopIndex;
}
codeElementsLine.AddParserDiagnostic(new ParserDiagnostic(ex.ToString(), start, stop,
line, null, code, ex));
}
CodeElement codeElement = codeElementBuilder.CodeElement;
if (codeElement != null)
{
// Attach consumed tokens and main document line numbers information to the code element
if (codeElement.ConsumedTokens.Count == 0)
{
// ISSUE #204:
var tempToken = tokenStream.Lt(1);
if (tempToken != null && tempToken != Token.END_OF_FILE)
{
// if not end of file,
// add next token to ConsumedTokens to know where is the CodeElement in error
codeElement.ConsumedTokens.Add((Token)tempToken);
// this alter CodeElements semantics: in addition to matched tokens,
// it includes the first token in error if no token has been matched
}
}
//TODO Issue #384 to discuss if this code should stay here:
//This should be in a Checker, but "codeElement.ConsumedTokens" is only set after all the checkers have been called
//Rule TCLIMITATION_NO_CE_ACROSS_SOURCES
if (codeElement.IsAcrossSourceFile())
{
DiagnosticUtils.AddError(codeElement,
"A Cobol statement cannot be across 2 sources files (eg. Main program and a COPY)",
MessageCode.TypeCobolParserLimitation);
}
// Add code element to the list
codeElementsLine.AddCodeElement(codeElement);
}
}
}
// If the parse tree contains errors
if (codeElementsParseTree.Diagnostics != null)
{
foreach (ParserDiagnostic d in codeElementsParseTree.Diagnostics)
{
if (d.OffendingSymbol != null)
{
CodeElementsLine codeElementsLine =
GetCodeElementsLineForToken((Token)d.OffendingSymbol);
if (codeElementsLine != null)
{
codeElementsLine.AddParserDiagnostic(d);
}
}
}
}
perfStatsForParserInvocation.OnStopTreeBuilding();
}
if (AntlrPerformanceProfiler != null)
{
AntlrPerformanceProfiler.EndParsingFile(cobolParser.ParseInfo.DecisionInfo, (int)(cobolParser.ParseInfo.GetTotalTimeInPrediction() / 1000000));
}
return(codeElementsLinesChanges);
}
/// <summary>
/// Incremental parsing of a set of processed tokens lines changes
/// </summary>
internal static IList <DocumentChange <ICodeElementsLine> > ParseProcessedTokensLinesChanges(TextSourceInfo textSourceInfo, ISearchableReadOnlyList <CodeElementsLine> documentLines, IList <DocumentChange <IProcessedTokensLine> > processedTokensLinesChanges, PrepareDocumentLineForUpdate prepareDocumentLineForUpdate, TypeCobolOptions compilerOptions)
{
// Collect all changes applied to the processed tokens lines during the incremental scan
IList <DocumentChange <ICodeElementsLine> > codeElementsLinesChanges = new List <DocumentChange <ICodeElementsLine> >();
// There are 2 reasons to re-parse a tokens line after a change :
// 1. The tokens line changed : these lines were already reset during the previous steps
// 2. If a tokens line that changed was involved in the parsing of a multiline code element, the whole group of lines must be parsed again
// --- PREPARATION PHASE : identify all parse sections where code elements need to be refreshed ---
IList <ParseSection> refreshParseSections = null;
// Iterate over all processed tokens changes detected by the PreprocessorStep :
// - refresh all the adjacent lines participating in a CodeElement
// - register the start and stop token for all sections of the document which need to be parsed again
if (processedTokensLinesChanges != null)
{
// If the document was cleared, everything must be parsed again
if (processedTokensLinesChanges[0].Type != DocumentChangeType.DocumentCleared)
{
refreshParseSections = new List <ParseSection>();
ParseSection lastParseSection = null;
foreach (DocumentChange <IProcessedTokensLine> tokensChange in processedTokensLinesChanges)
{
if (lastParseSection == null || tokensChange.LineIndex > lastParseSection.StopLineIndex)
{
lastParseSection = CheckIfAdjacentLinesNeedRefresh(tokensChange.Type, tokensChange.LineIndex, documentLines, prepareDocumentLineForUpdate, codeElementsLinesChanges, lastParseSection);
refreshParseSections.Add(lastParseSection);
}
}
}
}
// --- INITIALIZE ANTLR CodeElements parser ---
// Create a token iterator on top of pre-processed tokens lines
ITokensLinesIterator tokensIterator = ProcessedTokensDocument.GetProcessedTokensIterator(textSourceInfo, documentLines);
// Create an Antlr compatible token source on top of the token iterator
TokensLinesTokenSource tokenSource = new TokensLinesTokenSource(
textSourceInfo.Name,
tokensIterator);
// Init parser
TokensLinesTokenStream tokenStream = new TokensLinesTokenStream(tokenSource, Token.CHANNEL_SourceTokens);
CodeElementsParser cobolParser = new CodeElementsParser(tokenStream);
// REVERT TO STD PARSER ==> TracingCobolParser cobolParser = new TracingCobolParser(tokenStream);
// Customize error recovery strategy
IAntlrErrorStrategy cobolErrorStrategy = new CodeElementErrorStrategy();
cobolParser.ErrorHandler = cobolErrorStrategy;
// Register all parse errors in a list in memory
ParserDiagnosticErrorListener errorListener = new ParserDiagnosticErrorListener();
cobolParser.RemoveErrorListeners();
cobolParser.AddErrorListener(errorListener);
// Prepare to analyze the parse tree
ParseTreeWalker walker = new ParseTreeWalker();
CodeElementBuilder codeElementBuilder = new CodeElementBuilder();
codeElementBuilder.Dispatcher = new CodeElementDispatcher();
codeElementBuilder.Dispatcher.CreateListeners();
// --- INCREMENTAL PARSING ---
// In case of incremental parsing, parse only the code sections we need to refresh
IEnumerator <ParseSection> parseSectionsEnumerator = null;
ParseSection currentParseSection = null;
if (refreshParseSections != null)
{
// Get the first code section we need to refresh
parseSectionsEnumerator = refreshParseSections.GetEnumerator();
parseSectionsEnumerator.MoveNext();
currentParseSection = parseSectionsEnumerator.Current;
// Seek just before the next code element starting token
tokenStream.SeekToToken(currentParseSection.StartToken);
tokenStream.StartLookingForStopToken(currentParseSection.StopToken);
}
// Parse a list of code elements for each parse section while advancing in the underlying token stream
do
{
// Reset parsing error diagnostics
cobolErrorStrategy.Reset(cobolParser);
// Try to parse code elements :
// - starting with the current parse section Start token
// - ending with the current parse section Stop token
CodeElementsParser.CobolCodeElementsContext codeElementsParseTree = null;
try {
codeElementsParseTree = cobolParser.cobolCodeElements();
} catch (Exception e) {
var currentToken = (Token)cobolParser.CurrentToken;
CodeElementsLine codeElementsLine = GetCodeElementsLineForToken(currentToken);
codeElementsLine.AddParserDiagnostic(new TokenDiagnostic(MessageCode.ImplementationError, currentToken, currentToken.Line, e));
}
if (codeElementsParseTree != null)
{
// If the parse tree is not empty
if (codeElementsParseTree.codeElement() != null && codeElementsParseTree.codeElement().Length > 0)
{
// Analyze the parse tree for each code element
foreach (var codeElementParseTree in codeElementsParseTree.codeElement())
{
// Get the first line that was parsed
var tokenStart = (Token)codeElementParseTree.Start;
CodeElementsLine codeElementsLine = GetCodeElementsLineForToken(tokenStart);
// Register that this line was updated
// COMMENTED FOR THE SAKE OF PERFORMANCE -- SEE ISSUE #160
//int updatedLineIndex = documentLines.IndexOf(codeElementsLine, codeElementsLine.InitialLineIndex);
//codeElementsLinesChanges.Add(new DocumentChange<ICodeElementsLine>(DocumentChangeType.LineUpdated, updatedLineIndex, codeElementsLine));
codeElementsLinesChanges.Add(new DocumentChange <ICodeElementsLine>(DocumentChangeType.LineUpdated, codeElementsLine.InitialLineIndex, codeElementsLine));
// Visit the parse tree to build a first class object representing the code elements
try { walker.Walk(codeElementBuilder, codeElementParseTree); }
catch (Exception ex)
{
var code = MessageCode.ImplementationError;
int line = 0; int start = 0; int stop = 0;
if (codeElementsLine.SourceTokens != null && codeElementsLine.SourceTokens.Count > 0)
{
start = codeElementsLine.SourceTokens[0].StartIndex;
stop = codeElementsLine.SourceTokens[codeElementsLine.SourceTokens.Count - 1].StopIndex;
}
codeElementsLine.AddParserDiagnostic(new ParserDiagnostic(ex.ToString(), start, stop, line, null, code));
}
CodeElement codeElement = codeElementBuilder.CodeElement;
if (codeElement != null)
{
// Attach consumed tokens and main document line numbers information to the code element
if (codeElement.ConsumedTokens.Count == 0)
{ // ISSUE #204:
if (tokenStream.Lt(1) != null)
{ // if not end of file,
// add next token to ConsumedTokens to know where is the CodeElement in error
codeElement.ConsumedTokens.Add((Token)tokenStream.Lt(1));
// this alter CodeElements semantics: in addition to matched tokens,
// it includes the first token in error if no token has been matched
}
}
//TODO Issue #384 to discuss if this code should stay here:
//This should be in a Checker, but "codeElement.ConsumedTokens" is only set after all the checkers have been called
//Rule TCLIMITATION_NO_CE_ACROSS_SOURCES
if (codeElement.IsAcrossSourceFile())
{
DiagnosticUtils.AddError(codeElement, "A Cobol statement cannot be across 2 sources files (eg. Main program and a COPY)", MessageCode.TypeCobolParserLimitation);
}
// Add code element to the list
codeElementsLine.AddCodeElement(codeElement);
if (codeElement.Diagnostics != null)
{
foreach (Diagnostic d in codeElement.Diagnostics)
{
codeElementsLine.AddParserDiagnostic(d);
}
}
}
}
}
// If the parse tree contains errors
if (codeElementsParseTree.Diagnostics != null)
{
foreach (ParserDiagnostic d in codeElementsParseTree.Diagnostics)
{
if (d.OffendingSymbol != null)
{
CodeElementsLine codeElementsLine = GetCodeElementsLineForToken((Token)d.OffendingSymbol);
codeElementsLine.AddParserDiagnostic(d);
}
}
}
}
// In case of incremental parsing, directly jump to next parse section in the token stream
// Else, simply start parsing the next CodeElement beginning with the next token
if (currentParseSection != null)
{
// Adavance to the next ParseSection
if (parseSectionsEnumerator.MoveNext())
{
currentParseSection = parseSectionsEnumerator.Current;
tokenStream.SeekToToken(currentParseSection.StartToken);
tokenStream.StartLookingForStopToken(currentParseSection.StopToken);
}
// No more section to parse
else
{
break;
}
}
}while (tokenStream.La(1) >= 0);
return(codeElementsLinesChanges);
}
