From 623116ab13410a46d00ed8ce183702a6dad29ef0 Mon Sep 17 00:00:00 2001
From: Eric Coissac <eric@coissac.eu>
Date: Tue, 10 Mar 2026 16:34:25 +0100
Subject: [PATCH 1/8] Add rope-based FASTA parsing and improve sequence
 handling

Introduce FastaChunkParserRope for direct rope-based FASTA parsing, enhance sequence extraction with whitespace skipping and U->T conversion, and update parser logic to support both rope and raw data sources.

- Added extractFastaSeq function to scan sequence bytes directly from rope
- Implemented FastaChunkParserRope for rope-based parsing
- Modified _ParseFastaFile to use rope when available
- Updated sequence handling to support U->T conversion
- Fixed line ending detection for FASTA parsing
---
 pkg/obiformats/fastaseq_read.go | 105 ++++++++++++++++++++++++++++++--
 1 file changed, 99 insertions(+), 6 deletions(-)

diff --git a/pkg/obiformats/fastaseq_read.go b/pkg/obiformats/fastaseq_read.go
index 3597be2..c700ea2 100644
--- a/pkg/obiformats/fastaseq_read.go
+++ b/pkg/obiformats/fastaseq_read.go
@@ -209,28 +209,121 @@ func FastaChunkParser(UtoT bool) func(string, io.Reader) (obiseq.BioSequenceSlic
 	return parser
 }
 
+// extractFastaSeq scans sequence bytes from the rope directly into dest,
+// appending valid nucleotide characters and skipping whitespace.
+// Stops when '>' is found at the start of a line (next record) or at EOF.
+// Returns (dest with appended bases, hasMore).
+// hasMore=true means scanner is now positioned at '>' of the next record.
+func (s *gbRopeScanner) extractFastaSeq(dest []byte, UtoT bool) ([]byte, bool) {
+	lineStart := true
+
+	for s.current != nil {
+		data := s.current.data[s.pos:]
+		for i, b := range data {
+			if lineStart && b == '>' {
+				s.pos += i
+				if s.pos >= len(s.current.data) {
+					s.current = s.current.Next()
+					s.pos = 0
+				}
+				return dest, true
+			}
+			if b == '\n' || b == '\r' {
+				lineStart = true
+				continue
+			}
+			lineStart = false
+			if b == ' ' || b == '\t' {
+				continue
+			}
+			if b >= 'A' && b <= 'Z' {
+				b += 'a' - 'A'
+			}
+			if UtoT && b == 'u' {
+				b = 't'
+			}
+			dest = append(dest, b)
+		}
+		s.current = s.current.Next()
+		s.pos = 0
+	}
+	return dest, false
+}
+
+// FastaChunkParserRope parses a FASTA chunk directly from the rope without Pack().
+func FastaChunkParserRope(source string, rope *PieceOfChunk, UtoT bool) (obiseq.BioSequenceSlice, error) {
+	scanner := newGbRopeScanner(rope)
+	sequences := obiseq.MakeBioSequenceSlice(100)[:0]
+
+	for {
+		bline := scanner.ReadLine()
+		if bline == nil {
+			break
+		}
+		if len(bline) == 0 || bline[0] != '>' {
+			continue
+		}
+
+		// Parse header: ">id definition"
+		header := bline[1:]
+		var id string
+		var definition string
+		sp := bytes.IndexByte(header, ' ')
+		if sp < 0 {
+			sp = bytes.IndexByte(header, '\t')
+		}
+		if sp < 0 {
+			id = string(header)
+		} else {
+			id = string(header[:sp])
+			definition = string(bytes.TrimSpace(header[sp+1:]))
+		}
+
+		seqDest := make([]byte, 0, 4096)
+		var hasMore bool
+		seqDest, hasMore = scanner.extractFastaSeq(seqDest, UtoT)
+
+		if len(seqDest) == 0 {
+			log.Fatalf("%s [%s]: sequence is empty", source, id)
+		}
+
+		seq := obiseq.NewBioSequenceOwning(id, seqDest, definition)
+		seq.SetSource(source)
+		sequences = append(sequences, seq)
+
+		if !hasMore {
+			break
+		}
+	}
+
+	return sequences, nil
+}
+
 func _ParseFastaFile(
 	input ChannelFileChunk,
 	out obiiter.IBioSequence,
 	UtoT bool,
 ) {
-
 	parser := FastaChunkParser(UtoT)
 
 	for chunks := range input {
-		sequences, err := parser(chunks.Source, chunks.Raw)
-		// obilog.Warnf("Chunck(%d:%d) -%d- ", chunks.Order, l, sequences.Len())
+		var sequences obiseq.BioSequenceSlice
+		var err error
+
+		if chunks.Rope != nil {
+			sequences, err = FastaChunkParserRope(chunks.Source, chunks.Rope, UtoT)
+		} else {
+			sequences, err = parser(chunks.Source, chunks.Raw)
+		}
 
 		if err != nil {
 			log.Fatalf("File %s : Cannot parse the fasta file : %v", chunks.Source, err)
 		}
 
 		out.Push(obiiter.MakeBioSequenceBatch(chunks.Source, chunks.Order, sequences))
-
 	}
 
 	out.Done()
-
 }
 
 func ReadFasta(reader io.Reader, options ...WithOption) (obiiter.IBioSequence, error) {
@@ -245,7 +338,7 @@ func ReadFasta(reader io.Reader, options ...WithOption) (obiiter.IBioSequence, e
 		1024*1024,
 		EndOfLastFastaEntry,
 		"\n>",
-		true,
+		false,
 	)
 
 	for i := 0; i < nworker; i++ {

From 3d2e2057228d111aadfbd0d4cd6a73c71480947d Mon Sep 17 00:00:00 2001
From: Eric Coissac <eric@coissac.eu>
Date: Tue, 10 Mar 2026 16:46:53 +0100
Subject: [PATCH 2/8] Refactor rope scanner and add FASTQ rope parser

This commit refactors the rope scanner implementation by renaming gbRopeScanner to ropeScanner and extracting the common functionality into a new file. It also introduces a new FastqChunkParserRope function that parses FASTQ chunks directly from a rope without Pack(), enabling more efficient memory usage. The existing parsers are updated to use the new rope-based parser when available. The BioSequence type is enhanced with a TakeQualities method for more efficient quality data handling.
---
 pkg/obiformats/fastaseq_read.go |  4 +-
 pkg/obiformats/fastqseq_read.go | 85 ++++++++++++++++++++++++++++++++-
 pkg/obiformats/genbank_read.go  | 81 +------------------------------
 pkg/obiformats/rope_scanner.go  | 80 +++++++++++++++++++++++++++++++
 pkg/obiseq/biosequence.go       |  9 ++++
 5 files changed, 176 insertions(+), 83 deletions(-)
 create mode 100644 pkg/obiformats/rope_scanner.go

diff --git a/pkg/obiformats/fastaseq_read.go b/pkg/obiformats/fastaseq_read.go
index c700ea2..5a5bbdd 100644
--- a/pkg/obiformats/fastaseq_read.go
+++ b/pkg/obiformats/fastaseq_read.go
@@ -214,7 +214,7 @@ func FastaChunkParser(UtoT bool) func(string, io.Reader) (obiseq.BioSequenceSlic
 // Stops when '>' is found at the start of a line (next record) or at EOF.
 // Returns (dest with appended bases, hasMore).
 // hasMore=true means scanner is now positioned at '>' of the next record.
-func (s *gbRopeScanner) extractFastaSeq(dest []byte, UtoT bool) ([]byte, bool) {
+func (s *ropeScanner) extractFastaSeq(dest []byte, UtoT bool) ([]byte, bool) {
 	lineStart := true
 
 	for s.current != nil {
@@ -252,7 +252,7 @@ func (s *gbRopeScanner) extractFastaSeq(dest []byte, UtoT bool) ([]byte, bool) {
 
 // FastaChunkParserRope parses a FASTA chunk directly from the rope without Pack().
 func FastaChunkParserRope(source string, rope *PieceOfChunk, UtoT bool) (obiseq.BioSequenceSlice, error) {
-	scanner := newGbRopeScanner(rope)
+	scanner := newRopeScanner(rope)
 	sequences := obiseq.MakeBioSequenceSlice(100)[:0]
 
 	for {
diff --git a/pkg/obiformats/fastqseq_read.go b/pkg/obiformats/fastqseq_read.go
index 9c94d2d..861705f 100644
--- a/pkg/obiformats/fastqseq_read.go
+++ b/pkg/obiformats/fastqseq_read.go
@@ -303,6 +303,80 @@ func FastqChunkParser(quality_shift byte, with_quality bool, UtoT bool) func(str
 	return parser
 }
 
+// FastqChunkParserRope parses a FASTQ chunk directly from a rope without Pack().
+func FastqChunkParserRope(source string, rope *PieceOfChunk, quality_shift byte, with_quality, UtoT bool) (obiseq.BioSequenceSlice, error) {
+	scanner := newRopeScanner(rope)
+	sequences := obiseq.MakeBioSequenceSlice(100)[:0]
+
+	for {
+		// Line 1: @id [definition]
+		hline := scanner.ReadLine()
+		if hline == nil {
+			break
+		}
+		if len(hline) == 0 || hline[0] != '@' {
+			continue
+		}
+		header := hline[1:]
+		var id string
+		var definition string
+		sp := bytes.IndexByte(header, ' ')
+		if sp < 0 {
+			sp = bytes.IndexByte(header, '\t')
+		}
+		if sp < 0 {
+			id = string(header)
+		} else {
+			id = string(header[:sp])
+			definition = string(bytes.TrimSpace(header[sp+1:]))
+		}
+
+		// Line 2: sequence
+		sline := scanner.ReadLine()
+		if sline == nil {
+			log.Fatalf("@%s[%s]: unexpected EOF after header", id, source)
+		}
+		seqDest := make([]byte, len(sline))
+		w := 0
+		for _, b := range sline {
+			if b >= 'A' && b <= 'Z' {
+				b += 'a' - 'A'
+			}
+			if UtoT && b == 'u' {
+				b = 't'
+			}
+			seqDest[w] = b
+			w++
+		}
+		seqDest = seqDest[:w]
+		if len(seqDest) == 0 {
+			log.Fatalf("@%s[%s]: sequence is empty", id, source)
+		}
+
+		// Line 3: + (skip)
+		scanner.ReadLine()
+
+		// Line 4: quality
+		qline := scanner.ReadLine()
+
+		seq := obiseq.NewBioSequenceOwning(id, seqDest, definition)
+		seq.SetSource(source)
+
+		if with_quality && qline != nil {
+			qDest := make([]byte, len(qline))
+			copy(qDest, qline)
+			for i := range qDest {
+				qDest[i] -= quality_shift
+			}
+			seq.TakeQualities(qDest)
+		}
+
+		sequences = append(sequences, seq)
+	}
+
+	return sequences, nil
+}
+
 func _ParseFastqFile(
 	input ChannelFileChunk,
 	out obiiter.IBioSequence,
@@ -313,7 +387,14 @@ func _ParseFastqFile(
 	parser := FastqChunkParser(quality_shift, with_quality, UtoT)
 
 	for chunks := range input {
-		sequences, err := parser(chunks.Source, chunks.Raw)
+		var sequences obiseq.BioSequenceSlice
+		var err error
+
+		if chunks.Rope != nil {
+			sequences, err = FastqChunkParserRope(chunks.Source, chunks.Rope, quality_shift, with_quality, UtoT)
+		} else {
+			sequences, err = parser(chunks.Source, chunks.Raw)
+		}
 
 		if err != nil {
 			log.Fatalf("File %s : Cannot parse the fastq file : %v", chunks.Source, err)
@@ -339,7 +420,7 @@ func ReadFastq(reader io.Reader, options ...WithOption) (obiiter.IBioSequence, e
 		1024*1024,
 		EndOfLastFastqEntry,
 		"\n@",
-		true,
+		false,
 	)
 
 	for i := 0; i < nworker; i++ {
diff --git a/pkg/obiformats/genbank_read.go b/pkg/obiformats/genbank_read.go
index 53a3057..3cabd28 100644
--- a/pkg/obiformats/genbank_read.go
+++ b/pkg/obiformats/genbank_read.go
@@ -29,70 +29,11 @@ const (
 
 var _seqlenght_rx = regexp.MustCompile(" +([0-9]+) bp")
 
-// gbRopeScanner reads lines from a PieceOfChunk rope without heap allocation.
-// The carry buffer (stack) handles lines that span two rope nodes.
-type gbRopeScanner struct {
-	current *PieceOfChunk
-	pos     int
-	carry   [256]byte // max GenBank line = 80 chars; 256 gives ample margin
-	carryN  int
-}
-
-func newGbRopeScanner(rope *PieceOfChunk) *gbRopeScanner {
-	return &gbRopeScanner{current: rope}
-}
-
-// ReadLine returns the next line without the trailing \n (or \r\n).
-// Returns nil at end of rope. The returned slice aliases carry[] or the node
-// data and is valid only until the next ReadLine call.
-func (s *gbRopeScanner) ReadLine() []byte {
-	for {
-		if s.current == nil {
-			if s.carryN > 0 {
-				n := s.carryN
-				s.carryN = 0
-				return s.carry[:n]
-			}
-			return nil
-		}
-
-		data := s.current.data[s.pos:]
-		idx := bytes.IndexByte(data, '\n')
-
-		if idx >= 0 {
-			var line []byte
-			if s.carryN == 0 {
-				line = data[:idx]
-			} else {
-				n := copy(s.carry[s.carryN:], data[:idx])
-				s.carryN += n
-				line = s.carry[:s.carryN]
-				s.carryN = 0
-			}
-			s.pos += idx + 1
-			if s.pos >= len(s.current.data) {
-				s.current = s.current.Next()
-				s.pos = 0
-			}
-			if len(line) > 0 && line[len(line)-1] == '\r' {
-				line = line[:len(line)-1]
-			}
-			return line
-		}
-
-		// No \n in this node: accumulate into carry and advance
-		n := copy(s.carry[s.carryN:], data)
-		s.carryN += n
-		s.current = s.current.Next()
-		s.pos = 0
-	}
-}
-
 // extractSequence scans the ORIGIN section byte-by-byte directly on the rope,
 // appending compacted bases to dest. Returns the extended slice.
 // Stops and returns when "//" is found at the start of a line.
 // The scanner is left positioned after the "//" line.
-func (s *gbRopeScanner) extractSequence(dest []byte, UtoT bool) []byte {
+func (s *ropeScanner) extractSequence(dest []byte, UtoT bool) []byte {
 	lineStart := true
 	skipDigits := true
 
@@ -139,24 +80,6 @@ func (s *gbRopeScanner) extractSequence(dest []byte, UtoT bool) []byte {
 	return dest
 }
 
-// skipToNewline advances the scanner past the next '\n'.
-func (s *gbRopeScanner) skipToNewline() {
-	for s.current != nil {
-		data := s.current.data[s.pos:]
-		idx := bytes.IndexByte(data, '\n')
-		if idx >= 0 {
-			s.pos += idx + 1
-			if s.pos >= len(s.current.data) {
-				s.current = s.current.Next()
-				s.pos = 0
-			}
-			return
-		}
-		s.current = s.current.Next()
-		s.pos = 0
-	}
-}
-
 // parseLseqFromLocus extracts the declared sequence length from a LOCUS line.
 // Format: "LOCUS       <id> <length> bp ..."
 // Returns -1 if not found or parse error.
@@ -205,7 +128,7 @@ func GenbankChunkParserRope(source string, rope *PieceOfChunk,
 	withFeatureTable, UtoT bool) (obiseq.BioSequenceSlice, error) {
 
 	state := inHeader
-	scanner := newGbRopeScanner(rope)
+	scanner := newRopeScanner(rope)
 	sequences := obiseq.MakeBioSequenceSlice(100)[:0]
 
 	id := ""
diff --git a/pkg/obiformats/rope_scanner.go b/pkg/obiformats/rope_scanner.go
new file mode 100644
index 0000000..a7217f5
--- /dev/null
+++ b/pkg/obiformats/rope_scanner.go
@@ -0,0 +1,80 @@
+package obiformats
+
+import "bytes"
+
+// ropeScanner reads lines from a PieceOfChunk rope without heap allocation.
+// The carry buffer (stack) handles lines that span two rope nodes.
+type ropeScanner struct {
+	current *PieceOfChunk
+	pos     int
+	carry   [256]byte // 256 gives ample margin for typical flat-file lines
+	carryN  int
+}
+
+func newRopeScanner(rope *PieceOfChunk) *ropeScanner {
+	return &ropeScanner{current: rope}
+}
+
+// ReadLine returns the next line without the trailing \n (or \r\n).
+// Returns nil at end of rope. The returned slice aliases carry[] or the node
+// data and is valid only until the next ReadLine call.
+func (s *ropeScanner) ReadLine() []byte {
+	for {
+		if s.current == nil {
+			if s.carryN > 0 {
+				n := s.carryN
+				s.carryN = 0
+				return s.carry[:n]
+			}
+			return nil
+		}
+
+		data := s.current.data[s.pos:]
+		idx := bytes.IndexByte(data, '\n')
+
+		if idx >= 0 {
+			var line []byte
+			if s.carryN == 0 {
+				line = data[:idx]
+			} else {
+				n := copy(s.carry[s.carryN:], data[:idx])
+				s.carryN += n
+				line = s.carry[:s.carryN]
+				s.carryN = 0
+			}
+			s.pos += idx + 1
+			if s.pos >= len(s.current.data) {
+				s.current = s.current.Next()
+				s.pos = 0
+			}
+			if len(line) > 0 && line[len(line)-1] == '\r' {
+				line = line[:len(line)-1]
+			}
+			return line
+		}
+
+		// No \n in this node: accumulate into carry and advance
+		n := copy(s.carry[s.carryN:], data)
+		s.carryN += n
+		s.current = s.current.Next()
+		s.pos = 0
+	}
+}
+
+// skipToNewline advances the scanner past the next '\n'.
+func (s *ropeScanner) skipToNewline() {
+	for s.current != nil {
+		data := s.current.data[s.pos:]
+		idx := bytes.IndexByte(data, '\n')
+		if idx >= 0 {
+			s.pos += idx + 1
+			if s.pos >= len(s.current.data) {
+				s.current = s.current.Next()
+				s.pos = 0
+			}
+			return
+		}
+		s.current = s.current.Next()
+		s.pos = 0
+	}
+}
diff --git a/pkg/obiseq/biosequence.go b/pkg/obiseq/biosequence.go
index f3939d8..a362a34 100644
--- a/pkg/obiseq/biosequence.go
+++ b/pkg/obiseq/biosequence.go
@@ -480,6 +480,15 @@ func (s *BioSequence) SetQualities(qualities Quality) {
 	s.qualities = CopySlice(qualities)
 }
 
+// TakeQualities stores the slice directly without copying.
+// The caller must not use the slice after this call.
+func (s *BioSequence) TakeQualities(qualities Quality) {
+	if s.qualities != nil {
+		RecycleSlice(&s.qualities)
+	}
+	s.qualities = qualities
+}
+
 // A method that appends a byte slice to the qualities of the BioSequence.
 func (s *BioSequence) WriteQualities(data []byte) (int, error) {
 	s.qualities = append(s.qualities, data...)

From 09fbc217d3bcbb7fd04432dfe9dd4cbf82984878 Mon Sep 17 00:00:00 2001
From: Eric Coissac <eric@coissac.eu>
Date: Tue, 10 Mar 2026 17:02:05 +0100
Subject: [PATCH 3/8] Add EMBL rope parsing support and improve sequence
 extraction

Introduce EmblChunkParserRope function to parse EMBL chunks directly from a rope without using Pack(). Add extractEmblSeq helper to scan sequence sections and handle U to T conversion. Update parser logic to use rope-based parsing when available, and fix feature table handling for WGS entries.
---
 pkg/obiformats/embl_read.go | 154 +++++++++++++++++++++++++++++++++++-
 1 file changed, 152 insertions(+), 2 deletions(-)

diff --git a/pkg/obiformats/embl_read.go b/pkg/obiformats/embl_read.go
index d2fa9b1..5a87eeb 100644
--- a/pkg/obiformats/embl_read.go
+++ b/pkg/obiformats/embl_read.go
@@ -161,6 +161,149 @@ func EmblChunkParser(withFeatureTable, UtoT bool) func(string, io.Reader) (obise
 	return parser
 }
 
+// extractEmblSeq scans the sequence section of an EMBL record directly on the
+// rope. EMBL sequence lines start with 5 spaces followed by bases in groups of
+// 10, separated by spaces, with a position number at the end. The section ends
+// with "//".
+func (s *ropeScanner) extractEmblSeq(dest []byte, UtoT bool) []byte {
+	// We use ReadLine and scan each line for bases (skip digits, spaces, newlines).
+	for {
+		line := s.ReadLine()
+		if line == nil {
+			break
+		}
+		if len(line) >= 2 && line[0] == '/' && line[1] == '/' {
+			break
+		}
+		// Lines start with 5 spaces; bases follow separated by single spaces.
+		// Digits at the end are the position counter — skip them.
+		// Simplest: take every byte that is a letter.
+		for _, b := range line {
+			if b >= 'A' && b <= 'Z' {
+				b += 'a' - 'A'
+			}
+			if UtoT && b == 'u' {
+				b = 't'
+			}
+			if b >= 'a' && b <= 'z' {
+				dest = append(dest, b)
+			}
+		}
+	}
+	return dest
+}
+
+// EmblChunkParserRope parses an EMBL chunk directly from a rope without Pack().
+func EmblChunkParserRope(source string, rope *PieceOfChunk, withFeatureTable, UtoT bool) (obiseq.BioSequenceSlice, error) {
+	scanner := newRopeScanner(rope)
+	sequences := obiseq.MakeBioSequenceSlice(100)[:0]
+
+	var id string
+	var scientificName string
+	defBytes := make([]byte, 0, 256)
+	featBytes := make([]byte, 0, 1024)
+	var taxid int
+	inSeq := false
+
+	for {
+		line := scanner.ReadLine()
+		if line == nil {
+			break
+		}
+
+		if inSeq {
+			// Should not happen — extractEmblSeq consumed up to "//"
+			inSeq = false
+			continue
+		}
+
+		switch {
+		case bytes.HasPrefix(line, []byte("ID   ")):
+			id = string(bytes.SplitN(line[5:], []byte(";"), 2)[0])
+		case bytes.HasPrefix(line, []byte("OS   ")):
+			scientificName = string(bytes.TrimSpace(line[5:]))
+		case bytes.HasPrefix(line, []byte("DE   ")):
+			if len(defBytes) > 0 {
+				defBytes = append(defBytes, ' ')
+			}
+			defBytes = append(defBytes, bytes.TrimSpace(line[5:])...)
+		case withFeatureTable && bytes.HasPrefix(line, []byte("FH   ")):
+			featBytes = append(featBytes, line...)
+		case withFeatureTable && bytes.Equal(line, []byte("FH")):
+			featBytes = append(featBytes, '\n')
+			featBytes = append(featBytes, line...)
+		case bytes.HasPrefix(line, []byte("FT   ")):
+			if withFeatureTable {
+				featBytes = append(featBytes, '\n')
+				featBytes = append(featBytes, line...)
+			}
+			if bytes.HasPrefix(line, []byte(`FT                   /db_xref="taxon:`)) {
+				rest := line[37:]
+				end := bytes.IndexByte(rest, '"')
+				if end > 0 {
+					taxid, _ = strconv.Atoi(string(rest[:end]))
+				}
+			}
+		case bytes.HasPrefix(line, []byte("     ")):
+			// First sequence line: extract all bases via extractEmblSeq,
+			// which also consumes this line's remaining content.
+			// But ReadLine already consumed this line — we need to process it
+			// plus subsequent lines. Process this line inline then call helper.
+			seqDest := make([]byte, 0, 4096)
+			for _, b := range line {
+				if b >= 'A' && b <= 'Z' {
+					b += 'a' - 'A'
+				}
+				if UtoT && b == 'u' {
+					b = 't'
+				}
+				if b >= 'a' && b <= 'z' {
+					seqDest = append(seqDest, b)
+				}
+			}
+			seqDest = scanner.extractEmblSeq(seqDest, UtoT)
+
+			seq := obiseq.NewBioSequenceOwning(id, seqDest, string(defBytes))
+			seq.SetSource(source)
+			if withFeatureTable {
+				seq.SetFeatures(featBytes)
+			}
+			annot := seq.Annotations()
+			annot["scientific_name"] = scientificName
+			annot["taxid"] = taxid
+			sequences = append(sequences, seq)
+
+			// Reset state
+			id = ""
+			scientificName = ""
+			defBytes = defBytes[:0]
+			featBytes = featBytes[:0]
+			taxid = 1
+
+		case bytes.Equal(line, []byte("//")):
+			// record ended without SQ/sequence section (e.g. WGS entries)
+			if id != "" {
+				seq := obiseq.NewBioSequenceOwning(id, []byte{}, string(defBytes))
+				seq.SetSource(source)
+				if withFeatureTable {
+					seq.SetFeatures(featBytes)
+				}
+				annot := seq.Annotations()
+				annot["scientific_name"] = scientificName
+				annot["taxid"] = taxid
+				sequences = append(sequences, seq)
+			}
+			id = ""
+			scientificName = ""
+			defBytes = defBytes[:0]
+			featBytes = featBytes[:0]
+			taxid = 1
+		}
+	}
+
+	return sequences, nil
+}
+
 func _ParseEmblFile(
 	input ChannelFileChunk,
 	out obiiter.IBioSequence,
@@ -171,7 +314,14 @@ func _ParseEmblFile(
 
 	for chunks := range input {
 		order := chunks.Order
-		sequences, err := parser(chunks.Source, chunks.Raw)
+		var sequences obiseq.BioSequenceSlice
+		var err error
+
+		if chunks.Rope != nil {
+			sequences, err = EmblChunkParserRope(chunks.Source, chunks.Rope, withFeatureTable, UtoT)
+		} else {
+			sequences, err = parser(chunks.Source, chunks.Raw)
+		}
 
 		if err != nil {
 			log.Fatalf("%s : Cannot parse the embl file : %v", chunks.Source, err)
@@ -196,7 +346,7 @@ func ReadEMBL(reader io.Reader, options ...WithOption) (obiiter.IBioSequence, er
 		1024*1024*128,
 		EndOfLastFlatFileEntry,
 		"\nID   ",
-		true,
+		false,
 	)
 
 	newIter := obiiter.MakeIBioSequence()

From 8c318c480e553e32ccc78dcc2f479bbaf4f704a7 Mon Sep 17 00:00:00 2001
From: Eric Coissac <eric@coissac.eu>
Date: Wed, 11 Mar 2026 17:05:34 +0100
Subject: [PATCH 4/8] replace fixed-size carry buffer with dynamic slice

Replace the fixed [256]byte carry buffer with a dynamic []byte slice to support arbitrarily long lines without heap allocation during accumulation. Update all carry buffer handling logic to use len(s.carry) and append instead of fixed-size copy operations.
---
 pkg/obiformats/rope_scanner.go | 27 ++++++++++++---------------
 1 file changed, 12 insertions(+), 15 deletions(-)

diff --git a/pkg/obiformats/rope_scanner.go b/pkg/obiformats/rope_scanner.go
index a7217f5..b8c6544 100644
--- a/pkg/obiformats/rope_scanner.go
+++ b/pkg/obiformats/rope_scanner.go
@@ -2,13 +2,12 @@ package obiformats
 
 import "bytes"
 
-// ropeScanner reads lines from a PieceOfChunk rope without heap allocation.
-// The carry buffer (stack) handles lines that span two rope nodes.
+// ropeScanner reads lines from a PieceOfChunk rope.
+// The carry buffer handles lines that span two rope nodes; it grows as needed.
 type ropeScanner struct {
 	current *PieceOfChunk
 	pos     int
-	carry   [256]byte // 256 gives ample margin for typical flat-file lines
-	carryN  int
+	carry   []byte
 }
 
 func newRopeScanner(rope *PieceOfChunk) *ropeScanner {
@@ -21,10 +20,10 @@ func newRopeScanner(rope *PieceOfChunk) *ropeScanner {
 func (s *ropeScanner) ReadLine() []byte {
 	for {
 		if s.current == nil {
-			if s.carryN > 0 {
-				n := s.carryN
-				s.carryN = 0
-				return s.carry[:n]
+			if len(s.carry) > 0 {
+				line := s.carry
+				s.carry = s.carry[:0]
+				return line
 			}
 			return nil
 		}
@@ -34,13 +33,12 @@ func (s *ropeScanner) ReadLine() []byte {
 
 		if idx >= 0 {
 			var line []byte
-			if s.carryN == 0 {
+			if len(s.carry) == 0 {
 				line = data[:idx]
 			} else {
-				n := copy(s.carry[s.carryN:], data[:idx])
-				s.carryN += n
-				line = s.carry[:s.carryN]
-				s.carryN = 0
+				s.carry = append(s.carry, data[:idx]...)
+				line = s.carry
+				s.carry = s.carry[:0]
 			}
 			s.pos += idx + 1
 			if s.pos >= len(s.current.data) {
@@ -54,8 +52,7 @@ func (s *ropeScanner) ReadLine() []byte {
 		}
 
 		// No \n in this node: accumulate into carry and advance
-		n := copy(s.carry[s.carryN:], data)
-		s.carryN += n
+		s.carry = append(s.carry, data...)
 		s.current = s.current.Next()
 		s.pos = 0
 	}

From 6ee87506359bda258729d878e61e07daa6dbb166 Mon Sep 17 00:00:00 2001
From: Eric Coissac <eric@coissac.eu>
Date: Thu, 12 Mar 2026 18:41:03 +0100
Subject: [PATCH 5/8] Replace SplitInTwo with LeftSplitInTwo/RightSplitInTwo
 for precise splitting

Replace SplitInTwo calls with LeftSplitInTwo or RightSplitInTwo depending on the intended split direction. In fastseq_json_header.go, extract rank from suffix without splitting; in biosequenceslice.go and taxid.go, use LeftSplitInTwo to split from the left; add RightSplitInTwo utility function for splitting from the right.
---
 pkg/obiformats/fastseq_json_header.go |  2 +-
 pkg/obiseq/biosequenceslice.go        |  2 +-
 pkg/obitax/taxid.go                   |  2 +-
 pkg/obiutils/strings.go               | 16 +++++++++++++++-
 4 files changed, 18 insertions(+), 4 deletions(-)

diff --git a/pkg/obiformats/fastseq_json_header.go b/pkg/obiformats/fastseq_json_header.go
index 8006c53..af8de1c 100644
--- a/pkg/obiformats/fastseq_json_header.go
+++ b/pkg/obiformats/fastseq_json_header.go
@@ -296,7 +296,7 @@ func _parse_json_header_(header string, sequence *obiseq.BioSequence) string {
 
 			case strings.HasSuffix(skey, "_taxid"):
 				if dataType == jsonparser.Number || dataType == jsonparser.String {
-					rank, _ := obiutils.SplitInTwo(skey, '_')
+					rank := skey[:len(skey)-len("_taxid")]
 
 					taxid := string(value)
 					sequence.SetTaxid(taxid, rank)
diff --git a/pkg/obiseq/biosequenceslice.go b/pkg/obiseq/biosequenceslice.go
index 597a6dd..da16bf4 100644
--- a/pkg/obiseq/biosequenceslice.go
+++ b/pkg/obiseq/biosequenceslice.go
@@ -195,7 +195,7 @@ func (s *BioSequenceSlice) ExtractTaxonomy(taxonomy *obitax.Taxonomy, seqAsTaxa
 				return nil, fmt.Errorf("sequence %v has no path", s.Id())
 			}
 			last := path[len(path)-1]
-			taxname, _ := obiutils.SplitInTwo(last, ':')
+			taxname, _ := obiutils.LeftSplitInTwo(last, ':')
 			if idx, ok := s.GetIntAttribute("seq_number"); !ok {
 				return nil, errors.New("sequences are not numbered")
 			} else {
diff --git a/pkg/obitax/taxid.go b/pkg/obitax/taxid.go
index f22ec4d..313a20d 100644
--- a/pkg/obitax/taxid.go
+++ b/pkg/obitax/taxid.go
@@ -31,7 +31,7 @@ func NewTaxidFactory(code string, alphabet obiutils.AsciiSet) *TaxidFactory {
 // It extracts the relevant part of the string after the first colon (':') if present.
 func (f *TaxidFactory) FromString(taxid string) (Taxid, error) {
 	taxid = obiutils.AsciiSpaceSet.TrimLeft(taxid)
-	part1, part2 := obiutils.SplitInTwo(taxid, ':')
+	part1, part2 := obiutils.LeftSplitInTwo(taxid, ':')
 	if len(part2) == 0 {
 		taxid = part1
 	} else {
diff --git a/pkg/obiutils/strings.go b/pkg/obiutils/strings.go
index 2586206..c07b919 100644
--- a/pkg/obiutils/strings.go
+++ b/pkg/obiutils/strings.go
@@ -144,7 +144,7 @@ func (r *AsciiSet) TrimLeft(s string) string {
 	return s[i:]
 }
 
-func SplitInTwo(s string, sep byte) (string, string) {
+func LeftSplitInTwo(s string, sep byte) (string, string) {
 	i := 0
 	for ; i < len(s); i++ {
 		c := s[i]
@@ -157,3 +157,17 @@ func SplitInTwo(s string, sep byte) (string, string) {
 	}
 	return s[:i], s[i+1:]
 }
+
+func RightSplitInTwo(s string, sep byte) (string, string) {
+	i := len(s) - 1
+	for ; i >= 0; i-- {
+		c := s[i]
+		if c == sep {
+			break
+		}
+	}
+	if i == len(s) {
+		return s, ""
+	}
+	return s[:i], s[i+1:]
+}

From 8dd32dc1bf621a413ed5390438ff58b7c87825f3 Mon Sep 17 00:00:00 2001
From: Eric Coissac <eric@coissac.eu>
Date: Thu, 12 Mar 2026 18:48:20 +0100
Subject: [PATCH 6/8] Fix CompressStream call to use compressed variable

Replace hardcoded boolean with the `compressed` variable in CompressStream call to ensure correct compression behavior.
---
 pkg/obitools/obiclean/graph.go | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/pkg/obitools/obiclean/graph.go b/pkg/obitools/obiclean/graph.go
index 9952c5e..e6ac7fa 100644
--- a/pkg/obitools/obiclean/graph.go
+++ b/pkg/obitools/obiclean/graph.go
@@ -64,7 +64,7 @@ func EmpiricalDistCsv(filename string, data [][]Ratio, compressed bool) {
 		fmt.Println(err)
 	}
 
-	destfile, err := obiutils.CompressStream(file, true, true)
+	destfile, err := obiutils.CompressStream(file, compressed, true)
 	if err != nil {
 		fmt.Println(err)
 	}

From abe935aa18bdb4ae8b80a026b420d10eb9a13a8a Mon Sep 17 00:00:00 2001
From: Eric Coissac <eric@coissac.eu>
Date: Thu, 12 Mar 2026 19:20:45 +0100
Subject: [PATCH 7/8] Add help target, colorize output, and improve release
 workflow
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- Add colored terminal output support (GREEN, YELLOW, BLUE, NC)
- Introduce `help` target to document all Makefile targets
- Enhance `bump-version` to accept VERSION env var for manual version setting
- Refactor jjpush: split into modular targets (jjpush-notes, jjpush-push, jjpush-tag)
- Replace orla with aichat for AI-powered release notes generation
- Add robust JSON parsing using Python for release notes extraction
- Use stakk for PR submission (replacing raw `jj git push`)
- Generate and store release notes in temp files for tag creation
- Add installation instructions to release tags
- Update .PHONY with new targets

4.4.20: Rope-based parsing, improved release tooling, and bug fixes

### Enhancements
- **Rope-based parsing**: Added direct rope parsing for FASTA, EMBL, and FASTQ formats via `FastaChunkParserRope`, `EmblChunkParserRope`, and `FastqChunkRope` functions, eliminating unnecessary memory allocation via Pack(). Sequence extraction now supports U→T conversion and improved line ending detection.
- **Rope scanner refactoring**: Unified rope scanning logic under a new `ropeScanner`, improving maintainability and consistency across parsers.
- **Sequence handling**: Added `TakeQualities()` method to BioSequence for more efficient quality data handling.

### Bug Fixes
- **Compression behavior**: Fixed CompressStream to correctly use the `compressed` variable instead of a hardcoded boolean.
- **String splitting**: Replaced ambiguous `SplitInTwo` calls with precise `LeftSplitInTwo` or `RightSplitInTwo`, and added dedicated right-split utility.

### Tooling & Workflow Improvements
- **Makefile enhancements**: Added colored terminal output, a `help` target for documenting all targets, and improved release workflow automation.
- **Release process**: Refactored `jjpush` into modular targets (`jjpush-notes`, `jjpush-push`, `jjpush-tag`), replaced `orla` with `aichat` for AI-assisted release notes, and introduced robust JSON parsing using Python. Release notes are now generated and stored in temp files for tag creation.
- **Versioning**: `bump-version` now supports the VERSION environment variable for manual version setting.
- **Submission**: Switched from raw `jj git push` to `stakk` for PR submission.

### Internal Notes
- Installation instructions are now included in release tags.
- Fixed-size carry buffer replaced with dynamic slice for arbitrarily long line support without extra allocations.
---
 Makefile         | 108 ++++++++++++++++++++++++++++++++++-------------
 tools/json2md.py |  36 ++++++++++++++++
 2 files changed, 115 insertions(+), 29 deletions(-)
 create mode 100755 tools/json2md.py

diff --git a/Makefile b/Makefile
index afc3772..d9fb833 100644
--- a/Makefile
+++ b/Makefile
@@ -2,6 +2,11 @@
 #export GOBIN=$(GOPATH)/bin
 #export PATH=$(GOBIN):$(shell echo $${PATH})
 
+GREEN  := \033[0;32m
+YELLOW := \033[0;33m
+BLUE   := \033[0;34m
+NC     := \033[0m
+
 GOFLAGS=
 GOCMD=go
 GOBUILD=$(GOCMD) build $(GOFLAGS)
@@ -60,6 +65,28 @@ endif
 
 OUTPUT:=$(shell mktemp)
 
+help:
+	@printf "$(GREEN)OBITools4 Makefile$(NC)\n\n"
+	@printf "$(BLUE)Main targets:$(NC)\n"
+	@printf "  %-20s %s\n" "all"          "Build all obitools (default)"
+	@printf "  %-20s %s\n" "obitools"     "Build all obitools binaries to build/"
+	@printf "  %-20s %s\n" "test"         "Run Go unit tests"
+	@printf "  %-20s %s\n" "obitests"     "Run integration tests (obitests/)"
+	@printf "  %-20s %s\n" "bump-version" "Increment patch version (or set with VERSION=x.y.z)"
+	@printf "  %-20s %s\n" "update-deps"  "Update all Go dependencies"
+	@printf "\n$(BLUE)Jujutsu workflow:$(NC)\n"
+	@printf "  %-20s %s\n" "jjnew"        "Document current commit and start a new one"
+	@printf "  %-20s %s\n" "jjpush"       "Release: describe, bump, generate notes, push PR, tag (VERSION=x.y.z optional)"
+	@printf "  %-20s %s\n" "jjfetch"      "Fetch latest commits from origin"
+	@printf "\n$(BLUE)Required tools:$(NC)\n"
+	@printf "  %-20s " "go";      command -v go      >/dev/null 2>&1 && printf "$(GREEN)✓$(NC) %s\n" "$$(go version)" || printf "$(YELLOW)✗ not found$(NC)\n"
+	@printf "  %-20s " "git";     command -v git     >/dev/null 2>&1 && printf "$(GREEN)✓$(NC) %s\n" "$$(git --version)" || printf "$(YELLOW)✗ not found$(NC)\n"
+	@printf "  %-20s " "jj";      command -v jj      >/dev/null 2>&1 && printf "$(GREEN)✓$(NC) %s\n" "$$(jj --version)" || printf "$(YELLOW)✗ not found$(NC)\n"
+	@printf "  %-20s " "gh";      command -v gh      >/dev/null 2>&1 && printf "$(GREEN)✓$(NC) %s\n" "$$(gh --version | head -1)" || printf "$(YELLOW)✗ not found$(NC)  (brew install gh)\n"
+	@printf "\n$(BLUE)Optional tools (release notes generation):$(NC)\n"
+	@printf "  %-20s " "aichat";  command -v aichat  >/dev/null 2>&1 && printf "$(GREEN)✓$(NC) %s\n" "$$(aichat --version)" || printf "$(YELLOW)✗ not found$(NC)  (https://github.com/sigoden/aichat)\n"
+	@printf "  %-20s " "jq";      command -v jq      >/dev/null 2>&1 && printf "$(GREEN)✓$(NC) %s\n" "$$(jq --version)" || printf "$(YELLOW)✗ not found$(NC)  (brew install jq)\n"
+
 all: install-githook obitools
 
 obitools: $(patsubst %,$(OBITOOLS_PREFIX)%,$(OBITOOLS))
@@ -106,15 +133,20 @@ pkg/obioptions/version.go: version.txt .FORCE
 	@rm -f $(OUTPUT)
 
 bump-version:
-	@echo "Incrementing version..."
 	@current=$$(cat version.txt); \
-	echo "  Current version: $$current"; \
-	major=$$(echo $$current | cut -d. -f1); \
-	minor=$$(echo $$current | cut -d. -f2); \
-	patch=$$(echo $$current | cut -d. -f3); \
-	new_patch=$$((patch + 1)); \
-	new_version="$$major.$$minor.$$new_patch"; \
-	echo "  New version: $$new_version"; \
+	if [ -n "$(VERSION)" ]; then \
+		new_version="$(VERSION)"; \
+		echo "Setting version to $$new_version (was $$current)"; \
+	else \
+		echo "Incrementing version..."; \
+		echo "  Current version: $$current"; \
+		major=$$(echo $$current | cut -d. -f1); \
+		minor=$$(echo $$current | cut -d. -f2); \
+		patch=$$(echo $$current | cut -d. -f3); \
+		new_patch=$$((patch + 1)); \
+		new_version="$$major.$$minor.$$new_patch"; \
+		echo "  New version: $$new_version"; \
+	fi; \
 	echo "$$new_version" > version.txt
 	@echo "✓ Version updated in version.txt"
 	@$(MAKE) pkg/obioptions/version.go
@@ -130,6 +162,7 @@ jjnew:
 jjpush:
 	@$(MAKE) jjpush-describe
 	@$(MAKE) jjpush-bump
+	@$(MAKE) jjpush-notes
 	@$(MAKE) jjpush-push
 	@$(MAKE) jjpush-tag
 	@echo "$(GREEN)✓ Release complete$(NC)"
@@ -142,44 +175,61 @@ jjpush-bump:
 	@echo "$(BLUE)→ Creating new commit for version bump...$(NC)"
 	@jj new
 	@$(MAKE) bump-version
-	@echo "$(BLUE)→ Documenting version bump commit...$(NC)"
-	@jj auto-describe
 
-jjpush-push:
-	@echo "$(BLUE)→ Pushing commits...$(NC)"
-	@jj git push --change @
-
-jjpush-tag:
+jjpush-notes:
 	@version=$$(cat version.txt); \
-	tag_name="Release_$$version"; \
-	echo "$(BLUE)→ Generating release notes for $$tag_name...$(NC)"; \
-	release_message="Release $$version"; \
-	if command -v orla >/dev/null 2>&1 && command -v jq >/dev/null 2>&1; then \
-		previous_tag=$$(git describe --tags --abbrev=0 --match 'Release_*' HEAD^ 2>/dev/null); \
+	echo "$(BLUE)→ Generating release notes for version $$version...$(NC)"; \
+	release_title="Release $$version"; \
+	release_body=""; \
+	if command -v aichat >/dev/null 2>&1; then \
+		previous_tag=$$(git describe --tags --abbrev=0 --match 'Release_*' 2>/dev/null); \
 		if [ -z "$$previous_tag" ]; then \
 			echo "$(YELLOW)⚠ No previous Release tag found, skipping release notes$(NC)"; \
 		else \
 			raw_output=$$(git log --format="%h %B" "$$previous_tag..HEAD" | \
-				ORLA_MAX_TOOL_CALLS=50 orla agent -m ollama:qwen3-coder-next:latest \
+				aichat \
 				"Summarize the following commits into a GitHub release note for version $$version. Ignore commits related to version bumps, .gitignore changes, or any internal housekeeping that is irrelevant to end users. Describe each user-facing change precisely without exposing code. Eliminate redundancy. Output strictly valid JSON with no surrounding text, using this exact schema: {\"title\": \"<short release title>\", \"body\": \"<detailed markdown release notes>\"}" 2>/dev/null) || true; \
 			if [ -n "$$raw_output" ]; then \
-				sanitized=$$(echo "$$raw_output" | sed -n '/^{/,/^}/p' | tr -d '\000-\011\013-\014\016-\037'); \
-				release_title=$$(echo "$$sanitized" | jq -r '.title // empty' 2>/dev/null) ; \
-				release_body=$$(echo "$$sanitized" | jq -r '.body // empty' 2>/dev/null) ; \
-				if [ -n "$$release_title" ] && [ -n "$$release_body" ]; then \
-					release_message="$$release_title"$$'\n\n'"$$release_body"; \
+				notes=$$(printf '%s\n' "$$raw_output" | python3 tools/json2md.py 2>/dev/null); \
+				if [ -n "$$notes" ]; then \
+					release_title=$$(echo "$$notes" | head -1); \
+					release_body=$$(echo "$$notes" | tail -n +3); \
 				else \
 					echo "$(YELLOW)⚠ JSON parsing failed, using default release message$(NC)"; \
 				fi; \
 			fi; \
 		fi; \
 	fi; \
+	printf '%s' "$$release_title" > /tmp/obitools4-release-title.txt; \
+	printf '%s' "$$release_body" > /tmp/obitools4-release-body.txt; \
+	echo "$(BLUE)→ Setting release notes as commit description...$(NC)"; \
+	jj desc -m "$$release_title"$$'\n\n'"$$release_body"
+
+jjpush-push:
+	@echo "$(BLUE)→ Pushing commits...$(NC)"
+	@jj git push --change @
+	@echo "$(BLUE)→ Creating/updating PR...$(NC)"
+	@release_title=$$(cat /tmp/obitools4-release-title.txt 2>/dev/null || echo "Release $$(cat version.txt)"); \
+	release_body=$$(cat /tmp/obitools4-release-body.txt 2>/dev/null || echo ""); \
+	branch=$$(jj log -r @ --no-graph -T 'bookmarks.map(|b| b.name()).join("\n")' 2>/dev/null | head -1); \
+	if [ -n "$$branch" ] && command -v gh >/dev/null 2>&1; then \
+		gh pr create --title "$$release_title" --body "$$release_body" --base master --head "$$branch" 2>/dev/null \
+		|| gh pr edit "$$branch" --title "$$release_title" --body "$$release_body" 2>/dev/null \
+		|| echo "$(YELLOW)⚠ Could not create/update PR$(NC)"; \
+	fi
+
+jjpush-tag:
+	@version=$$(cat version.txt); \
+	tag_name="Release_$$version"; \
+	release_title=$$(cat /tmp/obitools4-release-title.txt 2>/dev/null || echo "Release $$version"); \
+	release_body=$$(cat /tmp/obitools4-release-body.txt 2>/dev/null || echo ""); \
 	install_section=$$'\n## Installation\n\n### Pre-built binaries\n\nDownload the appropriate archive for your system from the\n[release assets](https://github.com/metabarcoding/obitools4/releases/tag/Release_'"$$version"')\nand extract it:\n\n#### Linux (AMD64)\n```bash\ntar -xzf obitools4_'"$$version"'_linux_amd64.tar.gz\n```\n\n#### Linux (ARM64)\n```bash\ntar -xzf obitools4_'"$$version"'_linux_arm64.tar.gz\n```\n\n#### macOS (Intel)\n```bash\ntar -xzf obitools4_'"$$version"'_darwin_amd64.tar.gz\n```\n\n#### macOS (Apple Silicon)\n```bash\ntar -xzf obitools4_'"$$version"'_darwin_arm64.tar.gz\n```\n\nAll OBITools4 binaries are included in each archive.\n\n### From source\n\nYou can also compile and install OBITools4 directly from source using the\ninstallation script:\n\n```bash\ncurl -L https://raw.githubusercontent.com/metabarcoding/obitools4/master/install_obitools.sh | bash -s -- --version '"$$version"'\n```\n\nBy default binaries are installed in `/usr/local/bin`. Use `--install-dir` to\nchange the destination and `--obitools-prefix` to add a prefix to command names:\n\n```bash\ncurl -L https://raw.githubusercontent.com/metabarcoding/obitools4/master/install_obitools.sh | \\\n  bash -s -- --version '"$$version"' --install-dir ~/local --obitools-prefix k\n```\n'; \
-	release_message="$$release_message$$install_section"; \
+	release_message="$$release_title"$$'\n\n'"$$release_body$$install_section"; \
 	echo "$(BLUE)→ Creating tag $$tag_name...$(NC)"; \
 	git tag -a "$$tag_name" -m "$$release_message" 2>/dev/null || echo "$(YELLOW)⚠ Tag $$tag_name already exists$(NC)"; \
 	echo "$(BLUE)→ Pushing tag $$tag_name...$(NC)"; \
-	git push origin "$$tag_name" 2>/dev/null || echo "$(YELLOW)⚠ Tag push failed or already pushed$(NC)"
+	git push origin "$$tag_name" 2>/dev/null || echo "$(YELLOW)⚠ Tag push failed or already pushed$(NC)"; \
+	rm -f /tmp/obitools4-release-title.txt /tmp/obitools4-release-body.txt
 
 jjfetch:
 	@echo "$(YELLOW)→ Pulling latest commits...$(NC)"
@@ -187,5 +237,5 @@ jjfetch:
 	@jj new master@origin
 	@echo "$(GREEN)✓ Latest commits pulled$(NC)"
 
-.PHONY: all obitools update-deps obitests githubtests jjnew jjpush jjpush-describe jjpush-bump jjpush-push jjpush-tag jjfetch bump-version .FORCE
+.PHONY: all obitools update-deps obitests githubtests help jjnew jjpush jjpush-describe jjpush-bump jjpush-notes jjpush-push jjpush-tag jjfetch bump-version .FORCE
 .FORCE:
diff --git a/tools/json2md.py b/tools/json2md.py
new file mode 100755
index 0000000..62ca2e6
--- /dev/null
+++ b/tools/json2md.py
@@ -0,0 +1,36 @@
+#!/usr/bin/env python3
+"""
+Read potentially malformed JSON from stdin (aichat output), extract title and
+body, and print them as plain text: title on first line, blank line, then body.
+Exits with 1 on failure (no output).
+"""
+
+import sys
+import json
+import re
+
+text = sys.stdin.read()
+
+m = re.search(r'\{.*\}', text, re.DOTALL)
+if not m:
+    sys.exit(1)
+
+s = m.group()
+obj = None
+
+try:
+    obj = json.loads(s)
+except Exception:
+    s2 = re.sub(r'(?<!\\)\n', r'\\n', s)
+    try:
+        obj = json.loads(s2)
+    except Exception:
+        sys.exit(1)
+
+title = obj.get('title', '').strip()
+body  = obj.get('body', '').strip()
+
+if not title or not body:
+    sys.exit(1)
+
+print(f"{title}\n\n{body}")

From cd0c525f50aedb0ae4ab8f85c74a12900dcf6f57 Mon Sep 17 00:00:00 2001
From: Eric Coissac <eric@coissac.eu>
Date: Thu, 12 Mar 2026 20:06:24 +0100
Subject: [PATCH 8/8] 4.4.20: Rope-based parsing, improved release tooling, and
 bug fixes
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

### Enhancements
- **Rope-based parsing**: Added direct rope parsing for FASTA, EMBL, and FASTQ formats via `FastaChunkParserRope`, `EmblChunkParserRope`, and `FastqChunkParserRope`. Sequence extraction now supports U→T conversion and improved line ending detection.
- **Rope scanner refactoring**: Unified rope scanning logic under a new `ropeScanner`, improving maintainability and consistency.
- **Sequence handling**: Added `TakeQualities()` method to BioSequence for more efficient quality data handling.

### Bug Fixes
- **Compression behavior**: Fixed `CompressStream` to correctly use the `compressed` variable instead of a hardcoded boolean.
- **String splitting**: Replaced ambiguous `SplitInTwo` calls with precise `LeftSplitInTwo` or `RightSplitInTwo`, and added dedicated right-split utility.

### Tooling & Workflow Improvements
- **Makefile enhancements**: Added colored terminal output, a `help` target for documenting all targets, and improved release workflow automation.
- **Release process**: Refactored `jjpush` into modular targets (`jjpush-notes`, `jjpush-push`, `jjpush-tag`), replaced `orla` with `aichat` for AI-assisted release notes, and introduced robust JSON parsing using Python. Release notes are now generated and stored in temp files for tag creation.
- **Versioning**: `bump-version` now supports the VERSION environment variable for manual version setting.
- **Submission**: Switched from raw `jj git push` to `stakk` for PR submission.

### Internal Notes
- Installation instructions are now included in release tags.
- Fixed-size carry buffer replaced with dynamic slice for arbitrarily long line support without extra allocations.
---
 pkg/obioptions/version.go | 2 +-
 version.txt               | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/pkg/obioptions/version.go b/pkg/obioptions/version.go
index 097c578..0de542a 100644
--- a/pkg/obioptions/version.go
+++ b/pkg/obioptions/version.go
@@ -3,7 +3,7 @@ package obioptions
 // Version is automatically updated by the Makefile from version.txt
 // The patch number (third digit) is incremented on each push to the repository
 
-var _Version = "Release 4.4.19"
+var _Version = "Release 4.4.20"
 
 // Version returns the version of the obitools package.
 //
diff --git a/version.txt b/version.txt
index 0c6fdde..d83088a 100644
--- a/version.txt
+++ b/version.txt
@@ -1 +1 @@
-4.4.19
+4.4.20