Version 2.0.0

OEChem 2.0.0

New features

  • OEMolDatabase class added for providing fast random read-only access to all the file formats OEChem supports. Please see the new Molecular Database Handling chapter of the documentation for a more thorough description.

Warning

The meaning of the .smi file extension has changed in OEChem 2.0. The .smi file extension will now retain stereochemistry information. This could cause problems for systems that relied upon .smi to strip away stereo chemistry.

Warning

OEChem::OEConfBase::operator C * has been removed from the API. C++ users who used this API for direct coordinate access or derived their own classes from OEConfBase will need to change to the following different idiom.

  • OEChem::OEConfBase::operator C * has been removed in favor of the new OEConfBase::GetCoordsPtr methods. This allows implementations of OEConfBase to use alternative data storage for coordinates, e.g., 64-bit double precision floating point. Users of OEConfBase should not use OEConfBase::GetCoordsPtr directly, instead relying on the OEConstCoords and OEMutableCoords convenience classes.

  • Due to the above change, the following code will now fail to compile:

    OEMol mol;
for (OEIter<OEConfBase> conf = mol.GetConfs(); conf; ++conf)
  mol.NewConf(*conf);

    This code was not doing what most programmers thought it was doing. It was actually implicitly casting the OEConfBase & into a float * and then calling the version of NewConf that takes a float pointer. Users should change their code to the following:

    OEMol mol;
for (OEIter<OEConfBase> conf = mol.GetConfs(); conf; ++conf)
  mol.NewConf(conf);

  • The following methods that rely on OEConfBaseT<float, 2> and OEConfBaseT<double, 3> template instatiations were removed and replaced with OEMolBase versions:

Note

This is part of a larger effort to migrate away from the OEChem::OEConfBaseT template class in favor a pure abstract OEConfBase class to allow for conformer storage in formats other than 32-bit precision float. This will only affect C++ users of the toolkit, even then, users who only used OEConfBase as a typedef should not be affected.

Major bug fixes

  • The OEFormat::MDL V3000 file format was unable to read multiple molecules from the same file if they were all in the V3000 format. Note, this did not affect the OEFormat::SDF files containing V3000 as those files contain ‘$ $ $ $’ to delimit separate molecule records.
  • OEChem::OEReadHeader will no longer corrupt the stack by arbitrarily zeroing out bytes on the stack whenever no OEHeader record is found in the OEFormat::OEB file.
  • OEMatch::AddPair will no longer crash if OEMatch::Clear was previously called.
  • oemolthreadbase::PutMol will now destroy the pointer passed to it if the underlying buffer returns false, e.g., the buffer has already been closed.
  • OEAddMols, OESuppressHydrogens, and OEPerceiveSymmetry will no longer cause a stack overflow and crash when the molecule has a large number of atoms.

Minor bug fixes

  • Removed unimplemented OEChem::OEFormat::TDT file format.
  • Removed OEChem::OEFormat::BIN format after being deprecated for 10 years.
  • OEMolToSmiles no longer outputs the title of the molecule.
  • If OE3DToAtomStereo or OE3DToBondStereo throws a warning message during a call to OEMolToSmiles, the warning message will no longer erroneously say it is during a call to OEWriteMolecule.
  • OESetComment and OEGetComment have have been slightly optimized for speed. There is also a larger optimization for memory and file space in the .oeb format. Previously, setting the comment to an empty string would write superfluous data to the .oeb file.
  • OECopySDData and OECopyPDBData will no longer increase the memory consumption of the destination molecule whenever the source molecule does not contain any data. This was causing OEFormat::PDB files read into a OEMCMolBase to use more memory than necessary.
  • OEFormat::MOL2 parser will no longer create bonds with zero-order from “dummy bonds”. Zero-order bonds cause problems for many OEChem algorithms like OEKekulize.
  • OEFormat::MOL2 parser will now properly ignore lines between molecule records that start with the pound sign, “#”. Previously, these lines would cause OEChem to spew a lot of warnings and cause the parser to fail.

Documentation fixes

  • The chapter about InChI failures has been removed since most of the InChI failures were fixed in the last release, OEChem 1.9.3 in 2013.Oct.
  • Release notes section re-organized to make the current release more prominent.
  • Added documentation and code example for OESmartsLexReplace. The old function named SmartsLexReplace without the leading OE is considered deprecated and has been removed from the documentation.

OESystem 2.0.0

New features

  • OEStringJoin now accepts a parameter to determine if the output string should end with the selected delimiter.
  • OEStringTokenizeQuoted added a parameter for whether consecutive delimiters should be treated as a single delimiter. The default behavior is exactly the same as previous versions.
  • OEHalfFloat class added for storing floating point data as a 16-bit representation as specified by the IEEE 754-2008 standard to save on memory consumption and bandwidth.
  • Collections of OEHalfFloat objects, std::vector and arrays, can be attached to OEBase objects and round-tripped through .oeb files to save on disk size.
  • The performance of reading plain-old-data attached to OEBase objects from .oeb files has been improved by about %10.
  • OEErrorLevelToString free function added.
  • OEBFPosTEndian added for allowing 64-bit integers to be round-tripped to binary formats regardless of machine endian-ness.
  • OEWriteData added to allow easily writing any data type as binary.

Major bug fixes

Minor bug fixes

  • OEBitVectorNumWords and OEBitVectorNumBytes now take and return size_t.

  • OESystem::OEBinaryTagMaxLength constant added and set to 1024, the maximum length of an explicit string tag in the .oeb file format.

  • OEStringTokenizeQuoted will no longer treat a quote as the end quote of a field if it is escaped by another quote. Table: OEStringTokenizeQuoted Change demonstrates the change to support proper CSV parsing.

    OEStringTokenizeQuoted Change
    OEChem 1.x “foo””,bar”,blah "foo""``|``bar"``|``blah
    OEChem 2.0+ “foo””,bar”,blah "foo"",bar"``|``blah

  • The size of OEBitVector object has been increased from 12 bytes to 16 bytes on 64-bit machines. The size is still 8 bytes on a 32-bit machine.

Documentation fixes

OEPlatform 2.0.0

New features

  • OEAddLicenseData function added to parse a string as if it is an OpenEye license file and then license the current process with it.

Major bug fixes

  • OEPlatform::oeistream::size was incorrectly being truncated to 32-bits on 32-bit machines. This resulted in incorrect file sizes being reported for files over 4 gigabytes in size on 32-bit machines.
  • OEPlatform::oeifstream::tell would return incorrect values if called after OEPlatform::oeifstream::size and after some bytes were already read from the stream.

Minor bug fixes

  • OEPlatform::OEMallocaPtr::GetPtr added to allow the object to be explicitly converted to a pointer type.
  • OEMutex and OETryMutex destructors will now destroy themselves on pthread based systems. Destroying mutexes is optional according to the pthread standard, but helpful in debugging possible deadlocks.
  • OEMutex and OETryMutex no longer use gthreads, instead using pthreads, allowing for integration with non-GCC systems like libc++ on OSX.

OEGrid 1.4.5

Major bug fixes

  • OEGridFileType::Ascii file writer will no longer sometimes corrupt the stack and crash.
  • When reading a CCP4 file, the standard deviation stored in the CCP4 header is now used to normalize the file.
  • OENormalizeGrid now properly normalizes by sigma (not variance)
  • Writing CCP4 maps now uses the original map statistics when possible.
  • Rotated skew grids attached as generic data are now round-trippable when saved to OEB
  • Fixed a crash when interpolating grids where the rotation matrix inverts the target grids bounding box
  • MTZ files with more than 18 columns are now read properly
  • Fixed a memory leak in OESequenceAlignment.
  • The constructor for the predicate OEHasResidueNumber now takes an int rather than an unsigned int because residue numbers can be negative.