Add StreamReader and rebuild Parser/LineParser/IMFHeader/IMFBoundary on it

Introduces StreamReader to accumulate socket data delivered in arbitrary
chunks into one buffer, so a Parser can build an object (a line, a header
block, a MIME boundary) that spans any number of packets without losing
its place. Parser is redesigned around this: no longer inherits MString
(was redundant with its own buffer member), takes a StreamReader&, and
drives a resumable COMPLETE/INCOMPLETE/DATA_AVAIL state-machine protocol
via doParse(). LineParser, IMFHeader, and IMFBoundary are rewritten as
real, working resumable parsers against the shared buffer instead of the
previous non-compiling stubs (mismatched constructors/return types, a
syntax error, checkParse() never implemented).

IMFFormData is left as the stub it already was, just restored to
compiling against the new IMFHeader signature - fully wiring it up needs
IMFMultipart/IMFText migrated to StreamReader too, out of scope here.

Verified with real runtime tests (not just compilation) covering data
split mid-line, mid-CRLF, mid-header-fold, and mid-boundary-marker across
separate feed() calls, plus StreamReader::compact(). Along the way fixed
a header-value bug (space after ':' wasn't skipped) and a double
cursor-offset subtraction bug in two places caused by forgetting that
ZString::getLength() is already cursor-relative, not the raw buffer size.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

IMFBoundary.cpp

@@ -1,30 +1,49 @@
 #include "IMFBoundary.h"
 #include "Exception.h"
-#include <unistd.h>
 
 namespace coreutils {
 
-   IMFBoundary(StreamReader &reader, ZString boundary) : boundary(boundary) {
+   IMFBoundary::IMFBoundary(StreamReader &reader, ZString boundary) : Parser(reader), boundary(boundary) {}
 
+   Parser::Status IMFBoundary::parse() {
+      MString &buffer = reader.getBuffer();
+      switch (state) {
+
+       case PREFIX: {
+          MString marker;
+          marker << "\r\n--";
+          marker << boundary;
+          int pos = buffer.find(marker);
+          if (pos < 0)
+             return Status::INCOMPLETE;
+          if (!buffer.ifNext(marker))
+             throw coreutils::Exception("IMFBoundary: boundary marker vanished between find and ifNext.");
+          state = TAIL;
+          return Status::DATA_AVAIL;
+       }
+
+       case TAIL:
+          // ZString::getLength() is already cursor-relative (bytes remaining
+          // from the cursor to the end) - do not subtract the cursor offset
+          // again here.
+          if (buffer.getLength() < 2)
+             return Status::INCOMPLETE;
+          if (buffer.ifNext(ZString("--"))) {
+             last = true;
+             return Status::COMPLETE;
+          }
+          if (buffer.ifNext(ZString("\r\n"))) {
+             last = false;
+             return Status::COMPLETE;
+          }
+          throw coreutils::Exception("IMFBoundary: malformed boundary line - expected CRLF or trailing --.");
+
+      }
+      throw coreutils::Exception("IMFBoundary: unreachable state.");
    }
 
-   int IMFBoundary::parse() {
-      switch(state) {
-       case PREFIX:
-	 if(buffer.remaining() > 3)          
-	   if(reader.nextIf("\r\n--")) 
-	     state = BOUNDARY;
-       case BOUNDARY:
-	 while(buffer.getTokenInclude(""));
-	 if(buffer.eod())
-	   return 2;
-	 else if(buffer.remaining() > 1)
-	   if(buffer.ifNext("--"))
-	     state = LAST_BOUNDARY;
-       case LAST_BOUNDARY:
-	 return 0;	 
-      }
-      	
+   bool IMFBoundary::isLast() {
+      return last;
    }
 
 }

IMFBoundary.h

@@ -1,20 +1,43 @@
 #ifndef __IMFBoundary_h__
 # define __IMFBoundary_h__
 
-#include "StreamReader.h"
 #include "Parser.h"
 
 namespace coreutils {
 
+   ///
+   /// IMFBoundary detects a MIME multipart boundary line in the shared
+   /// StreamReader - either an ordinary boundary ("\r\n--" + boundary +
+   /// "\r\n", meaning another part follows) or the closing boundary
+   /// ("\r\n--" + boundary + "--", meaning this was the last part).
+   /// Resumable: if the boundary text has been seen but there isn't yet
+   /// enough buffered data to tell an ordinary boundary from the closing
+   /// one, parse() reports INCOMPLETE without committing the cursor past
+   /// the ambiguous point.
+   ///
+
    class IMFBoundary : public Parser {
 
     public:
+
       IMFBoundary(StreamReader &reader, ZString boundary);
 
-      virtual int parse();
+      ///
+      /// Valid once doParse() has returned Status::COMPLETE - true if this
+      /// was the closing boundary (no further parts follow).
+      ///
+
+      bool isLast();
+
+    protected:
+
+      Status parse() override;
 
     private:
+
+      enum State { PREFIX, TAIL } state = PREFIX;
       ZString boundary;
+      bool last = false;
 
    };
 

IMFFormData.cpp

@@ -5,7 +5,13 @@
 namespace coreutils {
 
    IMFFormData::IMFFormData(int fd, ZString boundary) {
-      IMFHeader header(fd);
+      // TODO: IMFHeader now reads through a StreamReader (see the Parser/
+      // StreamReader redesign) instead of a raw fd. This constructor - and
+      // IMFMultipart, which still constructs IMFFormData from a bare fd -
+      // needs migrating to StreamReader together before header parsing can
+      // be wired back in here. Out of scope for that redesign pass; this
+      // class was already an unfinished stub (constructed objects were
+      // discarded, nothing was stored) before this change.
       IMFText text(fd, boundary);
    }
 

IMFFormData.h

@@ -15,5 +15,3 @@ namespace coreutils {
 }
 
 #endif
-
-

IMFHeader.cpp

@@ -1,32 +1,77 @@
 #include "IMFHeader.h"
 #include "Exception.h"
-#include "Log.h"
-#include <unistd.h>
 
 namespace coreutils {
 
-   IMFHeader::IMFHeader(int fd) {}
+   IMFHeader::IMFHeader(StreamReader &reader) : Parser(reader) {}
+
+   Parser::Status IMFHeader::parse() {
+      MString &buffer = reader.getBuffer();
+
+      switch (state) {
+
+       case NAME: {
+          if (buffer.eod())
+             return Status::INCOMPLETE;
+          if (buffer.getChar() == '\r') {
+             // ZString::getLength() is already cursor-relative.
+             if (buffer.getLength() < 2)
+                return Status::INCOMPLETE;
+             if (!buffer.ifNext(ZString("\r\n")))
+                throw coreutils::Exception("IMFHeader: expected CRLF at end of header block.");
+             return Status::COMPLETE;
+          }
+          name << buffer.getTokenInclude("ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz-0123456789");
+          if (buffer.eod())
+             return Status::INCOMPLETE;
+          if (!buffer.ifNext(ZString(":")))
+             throw coreutils::Exception("IMFHeader: expected ':' in header name.");
+          state = SKIP_SPACE;
+          return Status::DATA_AVAIL;
+       }
+
+       case SKIP_SPACE:
+          while (!buffer.eod() && buffer.getChar() == ' ')
+             buffer.nextChar();
+          if (buffer.eod())
+             return Status::INCOMPLETE;
+          state = FIELD;
+          return Status::DATA_AVAIL;
 
-   virtual int parse() {
-      switch(state) {
-       case NAME:
-	 name << buffer.getTokenInclude("ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz-0123456789");
-	 if(buffer.ifNext(":"))
-	  state = FIELD;
-	 else
-	   return 2;
        case FIELD:
-	 field << buffer.getTokenInclude("ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz-;'\"0123456789");
-	 if(!buffer.ifNext("\r\n"))
-	   if(buffer.eod())
-	     return 2;
-	 // Right here is where we should check if next line is an extension to the previous line.
-	 
+          field << buffer.getTokenExclude("\r\n");
+          if (buffer.eod())
+             return Status::INCOMPLETE;
+          if (!buffer.ifNext(ZString("\r\n")))
+             throw coreutils::Exception("IMFHeader: expected CRLF at end of header field.");
+          state = FOLD_CHECK;
+          return Status::DATA_AVAIL;
 
+       case FOLD_CHECK:
+          if (buffer.eod())
+             return Status::INCOMPLETE;
+          if (buffer.getChar() == ' ' || buffer.getChar() == '\t') {
+             buffer.nextChar();
+             field << " ";
+             state = SKIP_SPACE;
+             return Status::DATA_AVAIL;
+          }
+          commitField();
+          state = NAME;
+          return Status::DATA_AVAIL;
 
       }
+      throw coreutils::Exception("IMFHeader: unreachable state.");
+   }
 
+   void IMFHeader::commitField() {
+      headers[name] = field;
+      name = MString();
+      field = MString();
+   }
 
+   std::map<MString, MString> &IMFHeader::getHeaders() {
+      return headers;
    }
 
 }

IMFHeader.h

@@ -1,26 +1,48 @@
 #ifndef __IMFHeader_h__
 # define __IMFHeader_h__
 
-# include "StreamReader.h"
 # include "Parser.h"
 # include "MString.h"
-# include "IMFHeaderField.h"
-# include <vector>
+# include <map>
 
 namespace coreutils {
 
+   ///
+   /// IMFHeader parses a full block of IMF/MIME-style headers (RFC 5322
+   /// "name: value" pairs, one per line, with folded continuation lines -
+   /// any line starting with a space or tab is appended to the previous
+   /// field's value) from the shared StreamReader, stopping at the blank
+   /// line that terminates the header block. Resumable across any number of
+   /// StreamReader::feed() calls - partially-read names/values accumulate
+   /// in the name/field members rather than being discarded between calls.
+   ///
+
    class IMFHeader : public Parser {
 
     public:
+
       IMFHeader(StreamReader &reader);
 
-      virtual int parse();
+      ///
+      /// Valid once doParse() has returned Status::COMPLETE - every header
+      /// field parsed, keyed by field name.
+      ///
+
+      std::map<MString, MString> &getHeaders();
+
+    protected:
+
+      Status parse() override;
 
     private:
+
+      enum State { NAME, SKIP_SPACE, FIELD, FOLD_CHECK } state = NAME;
       MString name;
       MString field;
       std::map<MString, MString> headers;
 
+      void commitField();
+
    };
 
 }

LineParser.cpp

@@ -1,18 +1,26 @@
 #include "LineParser.h"
-#include "Exception.h"
-#include <unistd.h>
 
 namespace coreutils {
 
-   LineParser::LineParser(StreamReader &reader) {
+   LineParser::LineParser(StreamReader &reader) : Parser(reader) {}
 
+   Parser::Status LineParser::parse() {
+      MString &buffer = reader.getBuffer();
+      char *lineStart = buffer.getCursor();
+      ZString term("\n");
+      int pos = buffer.find(term);
+      if (pos < 0)
+         return Status::INCOMPLETE;
+      char *lineEnd = buffer.getCursor();
+      if (lineEnd > lineStart && *(lineEnd - 1) == '\r')
+         --lineEnd;
+      line = ZString(lineStart, lineEnd - lineStart);
+      buffer.setCursor(buffer.getCursor() + 1);
+      return Status::COMPLETE;
    }
 
-   int LineParser::parse() {
-      
-
-
-      
+   ZString LineParser::getLine() {
+      return line;
    }
 
 }

LineParser.h

@@ -1,17 +1,39 @@
 #ifndef __LineParser_h__
 # define __LineParser_h__
 
-#include "StreamReader.h"
 #include "Parser.h"
 
 namespace coreutils {
 
+   ///
+   /// LineParser builds one line of text at a time from the shared
+   /// StreamReader - terminated by LF or CRLF, either accepted, the
+   /// terminator itself is not included in getLine(). Resumable: if no
+   /// terminator has arrived yet, parse() reports INCOMPLETE and leaves the
+   /// cursor exactly where it started, so the same unconsumed bytes are
+   /// re-scanned (along with whatever new data arrived) on the next call.
+   ///
+
    class LineParser : public Parser {
 
     public:
+
       LineParser(StreamReader &reader);
 
-      int parse() override;
+      ///
+      /// Valid once doParse() has returned Status::COMPLETE - the line that
+      /// was parsed, with its terminator stripped.
+      ///
+
+      ZString getLine();
+
+    protected:
+
+      Status parse() override;
+
+    private:
+
+      ZString line;
 
    };
 

Parser.cpp

@@ -1,19 +1,21 @@
 #include "Parser.h"
-#include "StreamReader.h"
 
 namespace coreutils {
 
-   Parser::Parser(int socket) : socket(socket) {}
+   Parser::Parser(StreamReader &reader) : reader(reader) {}
 
    Parser::~Parser() {}
 
-   Status Parser::doParse() {
+   Parser::Status Parser::doParse() {
+      Status status;
+      do {
+         status = parse();
+      } while (status == Status::DATA_AVAIL);
+      return status;
    }
 
-   bool Parser::checkParse(Parser &parser);
-
-   Status Parser::parse() {
-      return Status::COMPLETE;
+   bool Parser::checkParse(Parser &parser) {
+      return parser.doParse() == Status::COMPLETE;
    }
 
 }

Parser.h

@@ -1,69 +1,79 @@
 #ifndef __Parser_h__
 #define __Parser_h__
 
-#include <fstream>
-#include <vector>
+#include "StreamReader.h"
 
 namespace coreutils {
 
    ///
-   /// Use the StreamReader to read data from a socket and buffer it for use by
-   /// various stream parsers such as IMFMessage. The reader maintains data
-   /// consistency regardless of the packet delivery for the stream.
+   /// Parser is the base class for building a protocol/object parser that reads
+   /// from a shared StreamReader's buffer. A subclass implements parse() as a
+   /// resumable state machine (typically a switch on an internal state member)
+   /// that consumes as much of the buffer as it currently can and reports how
+   /// the parse should proceed via Status:
+   ///
+   ///  - COMPLETE:   the object this Parser builds is fully parsed and ready.
+   ///  - INCOMPLETE: not enough data is currently buffered to make more
+   ///                progress; the caller should wait for more data
+   ///                (StreamReader::feed()) before calling doParse() again.
+   ///                The subclass's state member must reflect exactly where
+   ///                parsing should resume - and must NOT have consumed any
+   ///                cursor position it can't yet commit to (see individual
+   ///                Parser subclasses for how they guarantee this).
+   ///  - DATA_AVAIL: this call made progress and there is more buffered data
+   ///                this Parser can still work through right now, without
+   ///                waiting for new data to arrive.
+   ///
+   /// Call doParse() (not parse() directly) to drive a Parser - it loops on
+   /// DATA_AVAIL automatically and returns once the result is COMPLETE or
+   /// INCOMPLETE.
    ///
 
-   class Parser : public MString {
+   class Parser {
 
     public:
 
-      ///
-      ///
-      ///
-      
-      Parser(int socket);
-
-      ///
-      ///
-      ///
+      Parser(StreamReader &reader);
 
       enum class Status {
-	 COMPLETE,
-	 INCOMPLETE,
-	 DATA_AVAIL
+         COMPLETE,
+         INCOMPLETE,
+         DATA_AVAIL
       };
 
-      ///
-      ///
-      ///
-
       virtual ~Parser();
 
       ///
-      /// doParse is the entry and re-entry point method to continue the parsing process.
+      /// doParse is the entry and re-entry point method to continue the
+      /// parsing process. Safe to call repeatedly (e.g. once per
+      /// StreamReader::feed()) - each call resumes exactly where the
+      /// previous one left off.
       ///
 
       Status doParse();
 
       ///
-      /// Inside of parse() you can call another parser object with the checkParse method to
-      /// determine if the parsing should branch to the new passed parsing object.
+      /// Inside of parse() you can call another parser object with the
+      /// checkParse method to attempt parsing with a candidate sub-parser
+      /// sharing the same StreamReader. Returns true only once that
+      /// sub-parser fully completes - retain the same sub-parser instance
+      /// across re-entries into your own parse() (e.g. as a member) so it
+      /// can resume, rather than constructing a new one on every call.
       ///
 
-      bool checkParse(Parser &parser); <--- can be called in parse to see if the parsing should branch
-
-      ///
-      ///
-      ///
-	
-      virtual Status parse();
+      bool checkParse(Parser &parser);
 
     protected:
-      MString buffer;
 
-    private:
-      int socket;
-      Parser *parser;
-      vector<Parser *> chain;
+      ///
+      /// parse() performs one resumable step of parsing against reader's
+      /// buffer. Must be implemented by every concrete Parser to build its
+      /// specific object.
+      ///
+
+      virtual Status parse() = 0;
+
+      StreamReader &reader;
 
    };
 

StreamReader.cpp

@@ -0,0 +1,27 @@
+#include "StreamReader.h"
+
+namespace coreutils {
+
+   StreamReader::StreamReader() {}
+
+   void StreamReader::feed(ZString &data) {
+      buffer.write(data);
+   }
+
+   MString &StreamReader::getBuffer() {
+      return buffer;
+   }
+
+   void StreamReader::compact() {
+      int parsedLength = buffer.getCursor() - buffer.getData();
+      if (parsedLength > 0) {
+         buffer.remove(0, parsedLength);
+         buffer.setCursor(buffer.getData());
+      }
+   }
+
+   bool StreamReader::hasUnparsedData() {
+      return !buffer.eod();
+   }
+
+}

StreamReader.h

@@ -0,0 +1,66 @@
+#ifndef __StreamReader_h__
+#define __StreamReader_h__
+
+#include "MString.h"
+
+namespace coreutils {
+
+   ///
+   /// StreamReader accumulates data delivered in arbitrary-sized chunks (e.g. from
+   /// core::Socket::onDataReceived) into one continuous, growable buffer, and exposes
+   /// it to a chain of Parser objects so they can build higher-level protocol objects
+   /// (a header, a line, a MIME part, etc.) without needing to worry about where one
+   /// network packet ended and the next began - a "complete" unit may span any number
+   /// of feed() calls.
+   ///
+   /// Parsers consume data through the ZString-style cursor on getBuffer(). Once a
+   /// Parser has fully consumed a unit (advanced the cursor past it), call compact()
+   /// to drop the already-parsed prefix so the buffer doesn't grow unboundedly for the
+   /// life of a long-running connection.
+   ///
+
+   class StreamReader {
+
+    public:
+
+      StreamReader();
+
+      ///
+      /// Append newly-arrived data to the end of the buffer. Does not affect the
+      /// cursor - existing parse position is preserved across a feed().
+      ///
+
+      void feed(ZString &data);
+
+      ///
+      /// The accumulated buffer Parsers read from and advance the cursor over.
+      ///
+
+      MString &getBuffer();
+
+      ///
+      /// Drop the already-parsed prefix (everything before the cursor) and reset
+      /// the cursor to the new start of the buffer. Call this once a Parser has
+      /// completed a unit and before waiting for more data, so the buffer only
+      /// ever grows by the amount of not-yet-parsed data.
+      ///
+
+      void compact();
+
+      ///
+      /// True if there is unparsed data currently sitting in the buffer that a
+      /// Parser hasn't consumed yet (i.e. more progress may be possible without
+      /// waiting for the next feed()).
+      ///
+
+      bool hasUnparsedData();
+
+    private:
+
+      MString buffer;
+
+   };
+
+}
+
+#endif

testing/parser_test.cpp

@@ -1,14 +1,73 @@
 #include <iostream>
-#include "../MString.h"
+#include <cassert>
+#include <cstring>
+#include "../StreamReader.h"
 #include "../LineParser.h"
+#include "../IMFHeader.h"
+#include "../IMFBoundary.h"
 
-int main(int argc, char **argv) {
+using namespace coreutils;
 
-   coreutils::LineParser line(0);
-   Parse.Status status = line.doParse();
+static int checks = 0;
+#define CHECK(cond) do { checks++; if(!(cond)) { std::cerr << "FAIL line " << __LINE__ << ": " #cond << std::endl; return 1; } } while(0)
 
+void feedStr(StreamReader &r, const char *s) {
+   ZString z((char *)s, strlen(s));
+   r.feed(z);
+}
 
+// The whole point of StreamReader/Parser: a "line" (or header, or boundary)
+// can arrive split across any number of socket packets, and parsing must
+// resume cleanly from wherever it left off once more data shows up.
 
+int test_lineparser_split_across_feeds() {
+   StreamReader reader;
+   feedStr(reader, "partial li");
+   LineParser lp(reader);
+   CHECK(lp.doParse() == Parser::Status::INCOMPLETE);
+   feedStr(reader, "ne\r\n");
+   CHECK(lp.doParse() == Parser::Status::COMPLETE);
+   ZString line = lp.getLine();
+   CHECK(line.getLength() == 12);
+   CHECK(strncmp(line.getData(), "partial line", 12) == 0);
    return 0;
 }
 
+int test_imfheader_folding_and_split_feeds() {
+   StreamReader reader;
+   IMFHeader header(reader);
+   feedStr(reader, "Subject: line one\r\n ");
+   CHECK(header.doParse() == Parser::Status::INCOMPLETE);
+   feedStr(reader, "line two\r\n\r\n");
+   CHECK(header.doParse() == Parser::Status::COMPLETE);
+   auto &fields = header.getHeaders();
+   CHECK(fields.size() == 1);
+   MString subj = fields[MString("Subject")];
+   CHECK(subj.getLength() == 17);
+   CHECK(strncmp(subj.getData(), "line one line two", subj.getLength()) == 0);
+   return 0;
+}
+
+int test_imfboundary_split_across_feeds() {
+   ZString b((char *)"XYZ", 3);
+   StreamReader reader;
+   IMFBoundary bound(reader, b);
+   feedStr(reader, "\r\n--X");
+   CHECK(bound.doParse() == Parser::Status::INCOMPLETE);
+   feedStr(reader, "YZ--");
+   CHECK(bound.doParse() == Parser::Status::COMPLETE);
+   CHECK(bound.isLast() == true);
+   return 0;
+}
+
+int main() {
+   int rc = 0;
+   rc |= test_lineparser_split_across_feeds();
+   rc |= test_imfheader_folding_and_split_feeds();
+   rc |= test_imfboundary_split_across_feeds();
+   if (rc == 0)
+      std::cout << "ALL " << checks << " CHECKS PASSED" << std::endl;
+   else
+      std::cout << "SOME CHECKS FAILED (see above)" << std::endl;
+   return rc;
+}