Hachoir project

Hachoir is a Python library used to represent of a binary file as a tree of Python objects. Each object has a type, a value, an address, etc. The goal is to be able to know the meaning of each bit in a file.

Why using slow Python code instead of fast hardcoded C code? Hachoir has many interesting features:

• Autofix: Hachoir is able to open invalid / truncated files
• Lazy: Open a file is very fast since no information is read from file, data are read and/or computed when the user ask for it
• Types: Hachoir has many predefined field types (integer, bit, string, etc.) and supports string with charset (ISO-8859-1, UTF-8, UTF-16, ...)
• Addresses and sizes are stored in bit, so flags are stored as classic fields
• Endian: You have to set endian once, and then number are converted in the right endian
• Editor: Using Hachoir representation of data, you can edit, insert, remove data and then save in a new file.

Website: http://bitbucket.org/haypo/hachoir/wiki/hachoir-core

Installation

For the installation, use setup.py or see: http://bitbucket.org/haypo/hachoir/wiki/Install

What's new in hachoir-core 1.2.1?

• Create configuration option "unicode_stdout" which avoid replacing stdout and stderr by objects supporting unicode string
• Create TimedeltaWin64 file type
• Support WINDOWS-1252 and WINDOWS-1253 charsets for GenericString
• guessBytesCharset() now supports ISO-8859-7 (greek)
• durationWin64() is now deprecated, use TimedeltaWin64 instead

What's new in hachoir-core 1.2?

• Create Field.getFieldType(): describe a field type and gives some useful informations (eg. the charset for a string)
• Create TimestampUnix64
• GenericString: only guess the charset once; if the charset attribute if not set, guess it when it's asked by the user.

What's new in hachoir-core 1.1?

Main change: string values are always encoded as Unicode. Details:

• Create guessBytesCharset() and guessStreamCharset()
• GenericString.createValue() is now always Unicode: if charset is not specified, try to guess it. Otherwise, use default charset (ISO-8859-1)
• RawBits: add createRawDisplay() to avoid slow down on huge fields
• Fix SeekableFieldSet.current_size (use offset and not current_max_size)
• GenericString: fix UTF-16-LE string with missing nul byte
• Add __nonzero__() method to GenericTimestamp
• All stream errors now inherit from StreamError (instead of HachoirError), and create and OutputStreamError
• humanDatetime(): strip microseconds by default (add optional argument to keep them)

What's new in hachoir-core 1.0?

Version 1.0.1 changelog:

• Rename parser.tags to parser.PARSER_TAGS to be compatible with future hachoir-parser 1.0

Visible changes:

• New field type: TimestampUUID60
• SeekableFieldSet: fix __getitem__() method and implement __iter__() and __len__() methods, so it can now be used in hachoir-wx
• String value is always Unicode, even on conversion error: use
• OutputStream: add readBytes() method
• Create Language class using ISO-639-2
• Add hachoir_core.profiler module to run a profiler on a function
• Add hachoir_core.timeout module to call a function with a timeout

Minor changes:

• Fix many spelling mistakes
• Dict: use iteritems() instead of items() for faster operations on huge dictionaries

What's new in hachoir-core 0.9?

Major changes:

• String value is ALWAYS unicode, even on charset conversion: use charset ISO-8859-1 on error
• Remove text_handler argument to bit and integer fields: use textHandler() instead
• Field.raw_display attribute creation is now fault tolerant and use cache
• SeekableFieldSet class is near complete and more stable
• GenericFieldSet: use a lock to avoid field creation recursion
• Add function limitedMemory() to call another function with a memory limit
• Add functions durationWin64(), timestampUUID60() and timedelta2seconds()
• Add "class" argument to SubFile constructor: specify a parser class

Changes:

• String classes: add truncate argument to constructor
• Rewrite humanDuration() which now accepts float and datetime.timedelta()
• BasicFieldSet: only use static_size if size is None (in constructor)
• Bit.raw_display is now "0" or "1" (and not "True" or "False")
• Float80.createValue() raises ValueError on overflow
• humanUnixAttributes() returns unicode string
• ISO 639: convert data to Unicode
• Field uses parent.nextFieldAddress() to get address and parent.getFieldIndex() to know its index (to be compatible with SeekableFieldSet)
• SeekableFieldSet: constructor use size argument
• SeekableFieldSet: add getField(), _getField(), __getitem()__, current_size and size attributes
• SeekableFieldSet: seekBit()/seekByte() checks address and field creation check bigger limit (address+size)
• String: support "CP037" charset

Bug fixes:

• GenericFieldSet: _addField() raise an ParserError if field type is invalid instead of an assertion
• SeekableFieldSet: fix seekBit(), getFieldByAddress() and done attribute
• GenericFieldSet: _fixFieldSize() doesn't display warning if the field is not deleted
• Fix createOrphanField(): always restore old field set address
• Fix Field._getDisplay(), cache wasn't used
• RawBytes: create method _createDisplay() to use classic prototype for createDisplay() method (no extra argument)
• Fix RawBytes._createDisplay() when string is truncated

What's new in hachoir-core 0.8?

New features:

• Field value and display attributes are fault tolerant
• New types: * Int24 and UInt24: signed/unsigned 24-bit integer ; * Float80: 80-bit floating point number ; * TimestampMSDOS32: 32-bit MS-DOS, since January 1st 1980 ; * TimestampUnix32: 32-bit UNIX, seconds since January 1st 1970 ; * TimestampMac32: 32-bit Mac, seconds since January 1st 1904 ; * TimestampWin64: 64-bit Windows, nanoseconds since January 1st 1600.
• Function createOrphanField(): allow to create a field at any address
• String: add "MacRoman" charset, and rename "UTF-16LE" to "UTF-16-LE" (and UTF-16BE to UTF-16-BE) for IronPython compatibility
• Write functions timestampUNIX(), timestampMac32(), timestampWin64(), and humanDatetime() for IronPython compatibility. Functions use UTC and not local timezone
• Add methods getSubIStream() and setSubIStream() to Field class

Other changes:

• Split GenericFieldSet into BasicFieldSet and GenericFieldSet, and create SeekableFieldSet (not working yet) class
• Remove EncodedField (replaced by SubFile).
• Move hachoir_core.editor to new subproject hachoir_editor
• Use ASCII and not ISO-8859-1 charset for raw display
• Field class inherits from Logger to have info(), warning() and error() methods

What's new in hachoir-core 0.7.2?

New features:

• Fault tolerant: create HACHOIR_ERRORS constant, a list of minor errors that can be ignored. This list is used in try/except to catch errors when creating field description or new field.
• setup.py only uses setuptools when it's asked
• GenericString: add truncate optional argument
• hexadecimal() text handler now accepts field of any size: align size to 4 bits

Bugfixes:

• HachoirError: use makePrintable() to convert string to Unicode (when needed)
• GenericInteger: raise error when field size is bigger than 256 bits
• Fix humanDuration() for duration bigger than one year
• Remove unused function align_nearest()

What's new in hachoir-core 0.7.1?

• New field type Float80 (80-bit flotting point number)
• New field type CompressedField (for compressed content)
• Create utility function createDict()
• Remove (old and unused) unit tests

What's new in hachoir-core 0.7?

Important changes:

• Rename the component "hachoir" to "hachoir-core"
• Editor supports Float32, Float64 and Character types
• Floats are now field set: it's possible to read sign, exponent and mantissa
• New type EncodedFile to parse "encoded" subfile: compressed, encrypted, encoded in base64, etc.
• New types GenericVector and UserVector to parse vectors
• Raw display is now closer to hexadecimal reprensentation for many types
• Rewrite documentation (hachoir-api.txt)
• Parser: don't have mime_type or tags attributes anymore
• Cleanup some Field methods: getOriginalDisplay() becomes raw_display property, methods _createValue() and _createDescription() become public, etc.
• FileInputStream() and FileOutputStream() now have optional "real_filename" to accept invalid unicode filename

Minor changes:

• Add __repr__() and __unicode__() methods to Field
• Use cache for array() method
• New types Int24 and UInt24
• Field value is now read-only
• FieldSet.seekBit()/seekByte() has null optional argument to create nul padding
• createPaddingField() raise an exception on invalid size instead of using assertion
• str2hex() always returns Unicode string
• Remove file export_xml.py (moved to hachoir-console component)

Bugfixes:

• Use fstat() to get input size instead of seek()+tell() since unusual files on /proc on Linux returns invalid size
• GenericString: fix UnixLine, fix UTF-16 with BOM, remove ISO-8859-12 charset (doesn't exist)
• Catch seek() error in InputStream and raise InputStreamError()
• Fix _fixFieldSize() method for field set with nul size
• Fix UnicodeStdout on 8-bit terminal (eg. MS-Dos terminal on Windows)
• Fix makePrintable(): quote quote character if needed

what's new in hachoir 0.6.1?

Bugfixes:

• Fix GenericString length attribute: wasn't initialized for UTF-* strings
• Fix and improve FakeArray (created with fieldset.array("name"))

Improvements:

• Add text_handler optional argument to Bits and RawBits
• On name duplicate error (a field was same name already exists), add "[]" to its name instead and display an error raising an exception

Minor changes:

• Add class documentation to all types (PaddingBits, Float32, ...)
• Fix reversed() code and __all__ constant in hachoir.compatibility
• Add available_types variable to hachoir.field (used by hachoir-wx)

What's new in Hachoir 0.6?

First of all, Hachoir project is split in backend: hachoir-core and hachoir-parser, and in frontends: hachoir-console, hachoir-urwid, hachoir-metadata, etc. Changes listed here are only about hachoir-core.

• Hachoir is able to edit a file: edit field value, delete field, insert new field
• Support of piped input: use small data cache and try to do the most without knowing stream size
• Better autofix feature to be able to open invalid / truncated files
• Support charset UTF-16 (and UTF-32) for strings
• Use Unicode strings everywhere (but only for text, not for binary data)
• Use gettext to translate messages => but disabled by default since setup.py is not able to compile .po file to .mo...

Changes for developers:

• Add new field types: NullBits, NullBytes and SubFile
• Create array() method for field set: self["name[%u]" % index] is now the same then self.array("name")[index]