Performance up

• Improved speed and memory use for reading channels with noDataLoading argument

Bug fixes and big files reading memory reduction

• fixed bug with duplicate long names in mdf3 and unsorted data

• for mdf4, read data in chunks to reduce memory use in case of big files

• Added compression option for HDF5 export

• changed from absolute to relative path import

• Reviewed minimum requirements for numpy to version 1.13.1

bug and performances improvements

• changed to objectify to parse xml instead of xpath
• implemented CANape specific partial conversion of text range table (not within spec)
• removed renaming of channels for recarray
• added possibility to read VLSD in DZ block

Bug fixes

Changed numbering to 3 digits in order to allow quick bug fix releases.

• Fixed issue with channel renaming
• Fixed wrong interpolation arguments
• Fixed bug for mdf3 range table conversion giving empty vector
• Fixed bug with mdf3 and empty comment
  removed bcolz depency because of its poor performance for vectors

Performance improvements

New features:

• cut method, to extract a part of the channels

Improvements:

• Reduced memory use and parsing time for files having lot of channels (>10k):
  -during data read, load info datagroup by datagroup and clean info after data block read in order to drastically reduce memory footprint for files having
  -converted channel object with many attributes into an empty object with many functions that is processing data from mdfinfo object

*Resample method: interpolation only for floating channels, not for integers.

Bug fixes:
-In case of unsorted data or canopen data, nodataloading argument flag was not working properly due to wrong channel namings.

Performance improvements

Added new features:

• In case of big file exploration, can load only file info and create dymmy mdf dict by using noDataLoading argument. Data is loaded by other methods when need transparently.
• To reduce memory consumption of big files, can conpress only the arrays
• Improved xml processing speed for mdf4 by using lxml and precompiled XPath.
• Improved performance when having lots of channels for read and write of blocks : refactored using more low level functions like pack/unpack
• fixed bugs related to embedded channels and big endian handling by dataRead
  Some pep8 improvements

Bug fixes

• Various bug fixes (especially regressions from recarray attribute name limitations fix)
• code refactorings
• handle duplicate channel names in same datablock among several channel group (unsorted data)

version 0.2.4

made mdfconverter compatible to PyQt5
introduced mdf3 unsorted read, to be confirmed as no sample file to test it
improved processing time for big number of channels, thanks to mattispasch contribution
implemented hidden and not aligned bytes handling in mdf3
changed from relative path import to absolute path for better robustess against platforms
fixed issue with dataRead import on windows by changing its location to module root
Fixed bug with special character channel (recarray is using python identifer having not allowed characters)

mfxz and array added functionnalities

Adds .mfxz reading capability (zip compressed .mfx=mdf4 file)
Improved composition implementation, especially for array

mdf 4.1 writing method introduction

several bug fixes but mainly introduces possibility to write mdf4 files.
By using mdf.write(), writes a file in the same major mdf version (3.x->3.3, 4.x-> 4.1) of the original.
If no file name given, original file name with '_new' appended is used.