Multistream WARCs and better Zstandard support

[tofuproxy.git] / warc / reader.go

diff --git a/warc/reader.go b/warc/reader.go
index e76bd51e299ceaffe2ba5e49cf98a552cb999a5f..06a123b4eece7fc352657a9d95491f6983518571 100644 (file)
--- a/warc/reader.go
+++ b/warc/reader.go
@@ -29,21 +29,22 @@ const CRLF = "\r\n"
 
 type Reader struct {
        Path    string
-       r       *bufio.Reader
-       rsc     io.ReadSeekCloser
+       rrr     RawRecordReader
+       br      *bufio.Reader
        offset  int64
        prevRec *Record
+       offsets []Offset
 }
 
 func NewReader(warcPath string) (*Reader, error) {
-       rsc, err := Open(warcPath)
+       rrr, err := Open(warcPath, nil, 0)
        if err != nil {
                return nil, err
        }
        return &Reader{
                Path: warcPath,
-               rsc:  rsc,
-               r:    bufio.NewReader(rsc),
+               rrr:  rrr,
+               br:   bufio.NewReader(rrr),
        }, nil
 }
 
@@ -51,12 +52,12 @@ func (r *Reader) next() error {
        if r.prevRec == nil {
                return nil
        }
-       if _, err := r.r.Discard(int(r.prevRec.Size)); err != nil {
+       if _, err := r.br.Discard(int(r.prevRec.Size)); err != nil {
                return err
        }
        r.offset += int64(r.prevRec.HdrLen) + r.prevRec.Size
        for i := 0; i < 2; i++ {
-               line, err := r.r.ReadString('\n')
+               line, err := r.br.ReadString('\n')
                if err != nil {
                        return err
                }
@@ -68,43 +69,53 @@ func (r *Reader) next() error {
        return nil
 }
 
-func (r *Reader) ReadRecord() (*Record, error) {
+func (r *Reader) ReadRecord() (*Record, io.Reader, error) {
        r.next()
-       line, err := r.r.ReadString('\n')
+       line, err := r.br.ReadString('\n')
        if err != nil {
-               return nil, err
+               return nil, nil, err
        }
        if !strings.HasPrefix(line, "WARC/") {
-               return nil, fmt.Errorf("non-WARC header: %q", line)
+               return nil, nil, fmt.Errorf("non-WARC header: %q", line)
        }
+       hdrLines := []string{line}
        hdrLen := len(line)
        hdr := NewHeader()
        for {
-               line, err := r.r.ReadString('\n')
+               line, err := r.br.ReadString('\n')
                if err != nil {
-                       return nil, err
+                       return nil, nil, err
                }
                hdrLen += len(line)
                if line == CRLF {
                        break
                }
+               hdrLines = append(hdrLines, line)
                hdr.AddLine(line)
        }
        size, err := strconv.ParseUint(hdr.Get("Content-Length"), 10, 64)
        if err != nil {
-               return nil, err
+               return nil, nil, err
        }
        rec := &Record{
                WARCPath: r.Path,
                Offset:   r.offset,
+               Size:     int64(size),
                Hdr:      hdr,
                HdrLen:   hdrLen,
-               Size:     int64(size),
+               HdrLines: hdrLines,
        }
        r.prevRec = rec
-       return rec, nil
+       return rec, &io.LimitedReader{R: r.br, N: int64(size)}, nil
+}
+
+func (r *Reader) RecordWasRead() {
+       r.prevRec.HdrLen = 0
+       r.prevRec.Size = 0
 }
 
 func (r *Reader) Close() error {
-       return r.rsc.Close()
+       err := r.rrr.Close()
+       r.offsets = r.rrr.Offsets()
+       return err
 }