New features

• HeaderValidator with WARC/1.1 standard ruleset
• ExtractTool: can now extract sequential concurrent records (--concurrent option)
• DedupeTool
  • In-memory cache for cross-URL digest-based deduplication (--cache-size option)
  • Now prints deduplication statistics (--dry-run and --quiet options)
  • Multi-threaded deduplication (--threads option)
• ValidateTool
  • Multi-threaded validation (--threads option)
• ParsingException message is now annotated with the source filename and record offset when available

Bugs fixed

• RFC5952 canonical form is now used for IPv6 addresses in WARC-IP-Address
• HttpParser in lenient mode now:
  • accepts responses missing version number
  • ignores header lines missing :
  • ignores folded status lines
• WarcParser: treats alexa/dat ARC records as not HTTP type

Bugs fixed

• Fixed URIs.parseLeniently() returning a different value to new URI() if the path was empty or the input contained percent encoded characters #90 #91
• Replaced some internal usages of record.targetURI() with record.target() to reduce the chance of runtime exceptions and preserve the exact original value

New features

• Added optional support for brotli content encoding #88
• Added HttpMessage.bodyDecoded() #88
• WarcTool: Added dedupe subcommand
• DedupeTool: Added --verbose option and silenced default logging

Bug fixes

• GunzipChannel: Fixed incorrect record length calculation when gzip footer aligns with the end of the buffer
• ValidateTool: Fixed digest validation #87
• DedupeTool: Used matchType=exact to properly handle CDX queries for URLs ending with *
• DedupeTool: Fixed record copying when transferTo copies fewer bytes than requested
• DedupeTool: Prevented appending of an empty gzip member when no records were deduplicated
• DedupeTool: Fixed exception when input files are in the current working directory

New features

• WarcReader and WarcParser gained a lenient parsing mode which:
  • permits ASCII control characters in header field names and values
  • allows lines to end with LF instead of CRLF
  • permits multi-digit WARC minor versions like "0.18"

New features

• Added MediaType.parseLeniently() and .isValid()

Changes

• Message.contentType() and other methods that internally call it now use the lenient MediaType parser instead of throwing IllegalArgumentException #83

Bugs fixed

• Improved compatibility with ARC variants (version-block length off by one, v2 version-block, spurious linefeeds) #82
• WarcParser: Context in parse error messages was incorrectly using the parser (file) position instead of buffer position

Bugs fixed

• Fixed ClosedChannelException when reading a WarcRevisit body after closing a previous one due to reuse of empty MessageBody. #80

Bugs fixed

• CDX formatting now percent encodes spaces, newlines and null characters in all string fields. This is non-standard but at least prevents us outputting invalid CDX lines.
• CdxRequestEncoder now handles requests with an invalid content-type header

Release 0.28.3

Bugs fixed:

• Fixed multithreading issue on GzipChannel write header #79

Changes:

• HttpRequest+HttpResponse in lenient mode now recovers when parsing the Content-Length header throws NumberFormatException
• WarcParser now tries to leniently parse ARC records containing corrupt dates