LibSBML news
libSBML NEWS -- History of user-visible changes
====================================================================
5.12.0 (2015-11-12)
====================================================================
STABLE RELEASE
* New features
- In SBML Level 2, some units were defined by default and did not
need to be explicitly included in a model. The
SBMLLevelVersionConverter adds these units explicitly when
converting to SBML Level 3. We have an added an option to
disable this functionality and return Level 3 models without
the units implied by Level 2.
- Sourceforge libSBML Tracker #242: 'addProduct etc using a species'
We have added functions addProduct/addReactant/addModifier that
take a Species as argument, along with a number of other optional
values to completely specify the SpeciesReference object being
created. This is a slight divergence from the direction libSBML
usually takes and we would be grateful for any feedback
on these functions.
- Validation speed has been dramatically improved for models that
make use of FunctionDefinitions and in particular nested
FunctionDefinitions.
- 'comp' package-specific updates:
- The 'flattening' routines have been expanded to correctly deal
with FBC V2.
- 'fbc' package-specific updates:
- The FBC Version 2 package has been accepted by the SBML Editors.
The code supporting this is now available as part of the stable
release.
* Bug fixes
- There was a very subtle bug in unit checking that incorrectly
applied a multiplication factor. Thanks to Mike Cooling for
the detailed example and anaylysis that revealed this. It
has now been fixed.
- In the new math code using the children of an ASTNode could result
in the math namespace declaration being unecessarily written out
in many places. This has been sorted out.
- Sourceforge libSBML Tracker #414: "A missing break statement" was
added.
* Configuration/build system changes
- The CMake files have added an option to dynamically link to a
python version on Mac OSX systems rather than automatically link.
Thanks to Kyle Medley for supplying the initial patch for this.
- The test code has been tweaked to build and run with MSVC 2015.
Fortunately the main code base worked with no issues.
* Miscellaneous
- Further documentation improvements have been added, especially for
the FBC package.
EXPERIMENTAL RELEASE
* New features
- 'multi' package-specific updates:
- Full validation has been added to the multi package code.
Thanks to Fengkai Zhang for his continued efforts on this.
====================================================================
5.11.6-experimental (2015-07-16)
====================================================================
*** Experimental only release ***
This is an experimental only release, that is, a release of those
files provided in the experimental folder of a full release.
This includes updates to all the SBML L3 packages that are not yet
accepted and undergoing development.
NOTE: An experimental release is not as rigorously tested as a
full release. Any changes/fixes listed as either referring to
stable or experimental code may change before the next full release.
STABLE RELEASE
* New features
- There is a new function on SBMLDocument, checkConsistencyWithStrictUnits(),
which applies the same validation as the checkConsistency() function with
one difference being that it will always apply unit validation using
exact identity comparison between units and it logs all unit
violations as Error regardless of the SBML Level and Version.
- There is a new converter that will convert to SBML L1V1. The 'regular'
SBMLLevelVersionConverter does not support this level and version
as we generally discourage the use of it. However users have requested
the facility and thus we have included a separate converter to meet
that need.
- 'const' versions of the getElementBySId() and getElementByMetaId()
functions have been added.
- 'fbc' package-specific updates:
- Code has been updated to use support the FBC V2 format that is currently
available as a Release Candidate document. Full validation for this
format has been included.
* Bug fixes
- The static function UnitDefinition::reorder failed to correctly deal
with situations where the same unit kind appeared multiple times. This
has been fixed. Thanks to Matthias König for reporting the problem.
- There were some discrepancies between math and math-legacy code reporting
the incorrect use of the MathML tag. This has been standardized
and more rigorously tested.
* Miscellaneous
- Several functions in libSBML checked for a NULL reference; something
that it is considered a well-formed program should not be able to
create. As more and more compilers are complaining about this code,
on the grounds it should not be possible to create the NULL reference,
we made the decision to remove this code. Please let us know if this
is an issue for you.
EXPERIMENTAL RELEASE
* New features
- L2V5 specific updates:
- A new validation rule was identified during the review of the SBML
Level 2 Version 5 specification. Support for this has been added.
- 'arrays' package-specific updates:
- The MathML elements included in the 'arrays' package has been
reduced. The code now reflects this.
- 'multi' package-specific updates:
- The code has been updated to reflect v 1.0.4 of the specification.
This includes full validation. Thanks to Fengkai Zhang for
contributing the code.
====================================================================
5.11.4 (2015-04-09)
====================================================================
STABLE RELEASE
* New features
- The ASTNode class has a new function double getValue() that
returns the numerical value of the ASTNode as a double. This
includes returning values for constants such as AST_CONSTANT_PI,
AST_CONSTANT_E, and AST_NAME_AVOGADRO, or '1' for an ASTNode of type
AST_CONSTANT_TRUE and '0' for AST_CONSTANT_FALSE. Note, it does
not attempt to evaluate the node and thus will return NaN for a
function or node representing a named variable, and AST_NAME_TIME.
- There has been a complete review of validation messages. These
have been unified in terms of spacing and line returns. In
addition many have been enhanced with additional model specific
information designed to assist in locating and fixing the issues.
- There is now an interface to PHP. This has been tested on all
platforms.
- Historically libSBML did not provide unsetFoo functions where
the attribute was a required attribute with a default value.
At user request we have changed this so every attribute has a
corresponding unset function. Note that where the attribute
has a default value it is still considered 'set' but not
considered to have been explicitly set and will therefore not
appear in the written output.
- The L3Parser will now correctly parse a unary plus.
- The L3Parser will now parse chains of relational operators.
Identical operators become ASTNodes with relational type and
each argument as a direct child. Thus 'x < y < z' becomes
an ASTNode of type AST_RELATIONAL_LT with three children
representing x, y and z. Non identical operators use a logical
'and' function to combine the arguments. Thus 'x < y > z'
becomes an ASTNode of type AST_LOGICAL_AND with two children
each representing the relational operator; that is
and(lt(x,y), gt(y,z))
- The ability to index lists in the python bindings has been improved.
- A new value 'LIBSBML_MISSING_METAID' has been added to the
OperationReturnValues_t enumeration. This value will be returned
when a user attempts to set an annotation on an object that has
no metaid attribute set. Note this is a change in behaviour as
previously the value returned would have been
'LIBSBML_UNEXPECTED_ATTRIBUTE'. We hope this change makes the
issue clearer to users.
- The package extension mechanism has been expanded to allow the
extension of individual ListOfFoo classes.
* Bug fixes
- Conversion from SBML Level 3 to Level 2 was losing some
LocalParameters. This has been sorted. Thanks to Matthias
König for spotting and reporting the problem.
- Fixed a bug where using the expat parser failed to correctly
report the presence of invalid XML namespaces.
- Sourceforge libSBML Tracker #393: LibSBML was failing when
built on an ARM processor running Fedora. This has been
sorted out. Thanks to Zbigniew Jedrzejewski-Szmek for
reporting the issue and testing solutions.
- The formulaParser was incorrectly parsing particular syntax
involving exponential, plus/minus signs and no spacing i.e.
'2e3-3' was failing to parse. This has been corrected.
Thanks to Kieran Smallbone for reporting this and apologies
for the length of time it took to fix.
- Sourceforge libSBML Tracker #402: The functions that deal with
user data on both ASTNode and SBase objects have been made
to work with const versions of objects where appropriate.
Thanks to Takeshi Abe for the request and the patch.
- 'comp' package-specific bug fixes:
- The code for determining whether an external file is available
has been further tightened up.
- 'fbc' package-specific bug fixes:
- A bug in the code that reads GeneAssociations as an annotation
meant information was being lost. This has been corrected.
Thanks to Matthias König for reporting the problem.
- 'layout' package-specific bug fixes:
- The names of some functions meant they were failing to
correctly port to the Python bindings. The binding mechanism
has been fixed to correct this. Thanks to Kieran Smallbone
for reporting this.
- The SpeciesReferenceGlyph attribute 'role' can take the value
'undefined' but this was not written out. This has been
changed.
* Configuration/build system changes
- This release includes pre-built binaries for the MATLAB
bindings that should work for many Linux and MacOSX
platforms. Please advise us it you find a combination
of platform and MATLAB version for which these fail.
* Miscellaneous
- The names of some arguments to functions have
been modified to convey the fact that the object
is not owned by the collection object to which it
has been added.
EXPERIMENTAL RELEASE
* New features
- 'distrib' package-specific updates:
- There are new example programs that will create the
examples given in the current specification.
- 'multi' package-specific updates:
- The multi code has been updated in line with the specification
Release 1.0.2 dated November 2014. This code has been contributed
by Fengkai Zhang. Many thanks, Fengkai.
- 'spatial' package-specific updates:
- The spatial code has been updated in line with the specification
Release 0.90 dated March 2015.
* Bug fixes
- 'multi' package-specific bug fixes:
- The multi attributes that can be added to a math element
where being lost when written out. This has been fixed.
====================================================================
5.11.2-experimental (2015-01-15)
====================================================================
********************************************************************
*** Experimental only release ***
This is an experimental only release, that is, a release of those
files provided in the experimental folder of a full release.
This includes updates to all the SBML L3 packages that are not yet
accepted and undergoing development.
NOTE: An experimental release is not as rigorously tested as a
full release. Any changes/fixes listed as either referring to
stable or experimental code may change before the next full release.
********************************************************************
STABLE RELEASE
* New features
- Functions have been added to allow users to determine the
number of SBML L3 packages within a document that the build
of libSBML does not support.
- The StripPackageConverter has a new option
'stripAllUnrecognized' which if set to true will strip all SBML
L3 packages that are not supported by the build of libSBML
being used.
- The XMLErrorLog::remove(unsigned int errorId) function only
removes the first instance of the error with the matching
identifier. A new function removeAll(unsigned int errorId)
has been added that will remove all instances of the error
with the given id.
* Bug fixes
- A stack overflow could occur when libSBML was built with certain
compilers on certain platforms. We have eliminated this problem
from the situations in which we can replicate it. The issue
occurred when performing intensive validation.
- The math-legacy code was being caught in an endless loop when
trying to read a nested piecewise function.
This has been fixed.
- Validation was failing to recognize some cases when the
conditional part of a piecewise function returned a boolean and
incorrectly indicated that such a model was incorrect. We have
fixed this as robustly as possible but encourage reports of
situations where we have missed something.
- 'layout' package-specific bug fixes:
- Validation of layout was not correctly reporting line
and column numbers. This has been corrected.
* Configuration/build system changes
- The CMake configuration has been improved to clearly
report failure to find files necessary to work with
particular language bindings.
- Sourceforge libSBML Tracker #394: The '\n' symbol
was not being correctly expanded on some systems. This
has been made more robust. Thanks to Antonio Trande
for reporting this.
EXPERIMENTAL RELEASE
* New features
- Support for the SBML Level 2 Version 5 Release 1
(Release Candidate) specification has been added.
Additions to libSBML include:
- new API for adding Nested Annotations used on
the existing CVTerm object:
int addNestedCVTerm(const CVTerm* term)
CVTerm* removeNestedCVTerm(unsigned int n)
unsigned int getNumNestedCVTerms()
CVTerm* getNestedCVTerm(unsigned int n)
List* getListNestdCVTerms()
- new and changed validation rules as in the specification
- the ability to convert core SBML to and from L2V5
- 'multi' package-specific updates:
- The multi code has been updated in line with the specification
Release 1.02 dated November 2014.
- 'spatial' package-specific updates:
- The spatial code has been updated in line with the specification
Release 0.89 dated November 2014.
* Bug fixes
- 'spatial' package-specific bug fixes:
- Sourceforge libsbml Tracker #387: The C# bindings for spatial
were not working as expected. These have been fixed. Thanks to
Harold Gómez for reporting this.
====================================================================
5.11.0 (2014-11-27)
====================================================================
STABLE RELEASE
* New features
- There is now an interface for JavaScript. This has been tested
on all platforms. Building requires SWIG version 3.0.2 or greater
and depends on node-gyp. Feedback on this would be greatly
appreciated.
- We have made a source package for the libSBML R interface
available for Linux and OSX users.
- The facility of using the data members of objects as attributes
in python has been extended to work with Python version 3.
- The SBMLDocument class now has an 'isSetModel()' function.
- There is a new writSBMLToStdString() function that returns a
std::string rather than a char * that has to be freed by the caller.
This facilitates cleaner memory management. Similarly there is
a writeMathMLToStdString() function.
- The errors reported by OutputSBML in the MATLAB binding have been
extended to provide more information about the specific field
causing a problem. Thanks to Thomas Pfau for providing a template
for adding this information.
- Support for the MIRIAM qualifiers hasTaxon, isInstanceOf and
hasInstance has been added.
- More information has been added to error messages.
- There are new functions to allow users to retrieve and print errors
from the ErrorLog based on the severity of the error. These are
SBMLError* getErrorWithSeverity(unsigned int n, unsigned int severity)
and printErrors(std::ostream& stream, unsigned int severity).
- 'comp' package-specific updates:
- A new flag has been added to the CompFlatteningConverter to
allow a user to specify a comment separated list of packages
that should be stripped before flattening is attempted.
- The CompFlatteningConverter has been made more rigorous
when it comes to specifying users options on models declared
as ExternalModelDefinitions.
* Bug fixes
- There was an issue with using the libsbml namespace and the
ReactionConverter. This has been fixed. Thanks to Andy Somogyi
for reporting it.
- The facility to use data members of objects as attributes
was broken. This has now been fixed.
- Code checking the values of units attributes was incorrectly
allowing the default Level 2 units, such as substance, to
be used in Level 3. This has been corrected.
- Using unicode filenames with MATLAB on a Windows system proves to
be very problematic. We have solved the problem but only for the
case where filenames are entered directly via the dialogs that
OutputSBML / TranslateSBML open up. There appears to be an issue
with MATLAB using such filenames directly from the workspace.
This has been added to the known issues documentation.
- The C API function ConversionProperties_removeOption now returns
a pointer to the option being removed. This is in line with other
remove functions.
- Validation failed to report an error when more than one model
element was encountered and merely returned a document with the
last model read. This has now been corrected.
- The getDerivedUnitDefinition function returns a pointer to an
existing unitDefinition object maintained by the model. However
in the case of a Parameter a new unitDefinition object was being
created, thus causing an inconsistency when it comes to freeing
memory. This has now been corrected and
Parameter::getDerivedUnitDefinition now returns an existing pointer.
- 'comp' package-specific bug fixes:
- Validation was failing to report when a portRef attribute
had been used on a Port object and indeed this situation caused
validation to crash in some cases. This has now been sorted.
* Configuration/build system changes
- When javac fails to compile a class no error was reported
and a useless jar file was created. The build has been
changed to highlight the issue.
- Sourceforge libsbml Tracker #380: The use of #include Windows.h
has been changed to lowercase to facilitate building using
MinGW. Thanks to Takeshi Abi for reporting the problem.
* Miscellaneous
- We have continued our review of code and many more memory leaks
have been plugged. Thanks to people, in particular Andy Somogyi,
for reporting these. Stable code has been rigorously tested using
both Valgrind (http://valgrind.org/) and Visual Leak Detector for
Visual C++ (https://vld.codeplex.com/). Experimental code is
currently being reviewed.
- Sourceforge libsbml Tracker #372: Documentation has been corrected.
Thanks to Gerard Weatherby for reporting the issues.
- The sample Java programs and configuration/importing instructions
for Java now use a shortened and simplified static loading block
(i.e., the part that explains how to use System.loadLibrary() to
load libSBML).
- We've made efforts to simplify the instructions for downloading,
configuring and accessing libSBML. Among the changes are that the
API language manuals for Java, Python and C# now focus their
instructions more on just those languages, rather than giving long
generic instructions for all possible languages.
- The SBML package extension mechanisms are now more thoroughly
documented.
EXPERIMENTAL RELEASE
* New features
Code for the Dynamic Structures ('dyn') package has been added.
This reflects the draft specification Release 0.1 dated
October 1 2014.
- 'distrib' package-specific updates:
- The UncertMLNode now uses a std::vector to save child members
rather than the List class.
- 'spatial' package-specific updates:
- The spatial code has been updated in line with the specification
Release 0.88 dated September 2014.
* Bug fixes
- 'distrib' package-specific bug fixes:
- A number of memory leaks have been plugged.
Thanks to Andy Somogyi for reporting them.
- 'groups' package-specific bug fixes:
- The values of the enumeration of SBMLGroupsTypeCode_t clashed
with the values used by the comp enumeration. The ones in
the groups package have been changed.
- 'req' package-specific bug fixes:
- Sourceforge libsbml Tracker #373: The code was not keeping
proper track of parent-child relationships which could
lead to incorrect namespaces being used. This is now sorted.
Thanks to Gerard Weatherby for reporting it.
====================================================================
5.10.2 (2014-07-17)
====================================================================
STABLE RELEASE
* New features
- The static functions implemented by the SBMLTransforms class
that are used by the converters that in-line FunctionDefinitions
or expand InitialAssignments has been made available to all language
bindings. These are: replaceFD, expandInitialAssignments
and evaluateASTNode. Examples of using these functions
are also provided.
- The L3ParserSettings object can now be configured to compare
built-in functions and constants caselessly (as it did in the
past; the default), or to only accept the lower-case forms.
- 'layout' package-specific updates:
- The C API has been improved and appropriate argument checking
has been implemented.
* Bug fixes
- There was an issue with detecting NULL elements with the
R bindings. This turns out to be a consequence of SWIG. A
bug report has been filed with the SWIG development team,
in the meantime, a revised SWIG file is included with the
libSBML source and is used by the libSBML build when
creating the R bindings.
- The unit consistency checking code was ignoring numerical
factors when checking the units of a KineticLaw matched the
model wide units of extent per time (substance per time in
SBML L1/L2). This has been changed. Thanks to Daniel Palm
for the report that exposed this.
- Unit checking code determines whether a math expression contains
parameters or numbers with undefined units. However, depending
on where these occurred within the expression, libSBML was
not always correctly establishing this. This has been fixed.
Thanks to Daniel Palm for reporting this.
- The unit consistency checking code was losing numerical factors
that were left after cancelling out units. This could affect
the overall definition of the remaining unit and thus impact code
that looked for exact identity between units. This has been fixed.
Thanks to Daniel Palm for providing an example of a case where this
was happening.
- There was a bug in writing out MathML that manifested itself
using the in-lining functionDefinition converter for a narrow
set of functions. This has been fixed.
- Validation has been improved so as to no longer warn a user that
a variable has no initial value if there exists an initialAssignment
or assignmentRule that would set the value.
- The SBMLError reporting an over-determined model was being logged
with the incorrect category of error. This has been changed.
- LibSBML incorrectly reported the units of 'pi' as radians. This
has been corrected to report 'dimensionless'. Thanks to Matthias
König for reporting this.
- The L3 parser did not recognize 'arcsinh' or 'asinh' as valid
built-in functions. This has been fixed.
- 'comp' package-specific bug fixes:
- When libSBML detects an SBML L3 package that it does not support
validation of 'comp' references can report that a reference may
be to an identifier within the unknown package. However, the code
was failing to check whether the reference could be located within
the xml relating to supported packages before issuing this warning.
This has been fixed. Thanks to Chris Myers for reporting it.
* Configuration/build system changes
- The CMake build system will now produce an uninstall target.
* Miscellaneous
- Continue our review of code a number of potential problems were
highlighted during static analysis. These have been fixed.
- Code review has also eliminated a number of compiler warnings
raised by various different compilers on various different
systems.
- Various memory leaks have been plugged. Thanks to people,
in particular Andy Somogyi, for reporting these.
- The short error messages for the following error codes in
SBMLErrorTable.h had trailing periods, which was inconsistent with
most of the other messages; the periods have been removed: 20109,
99109, 99401, 99402, 99403, 99404, 99405, 99406.
- In the API documentation, the instructions for setting PYTHONPATH
was not correct for the installation produced by using CMake
instead of GNU configure. Further, the PYTHONPATH instructions
for Ubuntu were actually for other Linux/Unix variants and vice
versa. This has been fixed by making the instructions use a
single, more general approach that should work everywhere.
- The API documentation lacked information about using PyPI for
Python, and about how to install the R interface. Fixed.
- Various issues in the Python documentation have been fixed. For
example, in the table of member methods for each class in the API
documentation for Python, the brief descriptions were missing --
all the short member descriptions had the same text, "Python
method signature(s)". This was caused by a bug in our custom
libSBML documentation generation tools.
- The error code table in the documentation for SBMLError has been
updated.
EXPERIMENTAL RELEASE
* New features
- 'arrays' package-specific updates:
- The arrays code has been updated in line with the specification
dated June 19 2014. Changes include a change in attribute names
and a revised set of supported math constructs.
- Support for parsing from infix to ASTNodes and vice versa has
been added for the majority of the math constructs currently
listed within the arrays specification.
- 'req' package-specific updates:
- The req code has been updated in line with the specification
Release 0.5 dated 30 May 2013. Users should be aware that
this update represents a significant change in code to reflect
the change in structure of the package.
- 'spatial' package-specific updates:
- The spatial code has been updated in line with the specification
Release 0.87 dated June 2014. Users should be aware that
this update represents a significant change in code to reflect
the change in structure of the package.
* Bug fixes
- There was a potential stack-overflow issue with packages that
extend SBase directly. This has been fixed where appropriate.
Thanks to Chris Myers for reporting it.
====================================================================
5.10.0 (2014-04-10)
====================================================================
STABLE RELEASE
* New features
- New ASTNode infrastructure
The code underlying the ASTNode class has been refactored to
facilitate the extension of MathML by SBML L3 packages. The
API and behaviour has not changed except in places where
invalid MathML was being incorrectly parsed or created. It is
no longer possible to add a child to an ASTNode that represents
a number, for example.
Extensive testing has been done to ensure that the API and
correct behaviour of the ASTNode class remains unchanged and
unless you have derived your own classes from the ASTNode class
this change should not affect your implementation in any way.
** Please inform us if this turns out to be incorrect. We have
tried to anticipate and test use cases but we are aware that
it is possible that we have missed some.
In order to allow people to continue to use the original AST code
if they choose, the source code archives provide this code within
a 'math-legacy' directory and the libSBML configuration system
provides an option to build libSBML using this legacy code.
- General maintenance: we have done a number of sweeps of code to
identify and add such things as missing virtual functions. We
have also investigated memory leaks reported by users and have
done some extensive memory leak checking on the more intensive
functionality such as validation/comp flattening etc. However,
this remains an ongoing task.
- There is a new converter, the SBMLReactionConverter, which
will replace any Reactions within a Model with the equivalent
RateRules. An example of the use of this converter can be
found at examples/c++/convertReactions.cpp.
- There are new functions available on the Model class:
InitialAssignment* getInitialAssignmentBySymbol(const std::string& symbol);
Rule* getRuleByVariable(const std::string& variable);
AssignmentRule* getAssignmentRuleByVariable(const std::string& variable);
RateRule* getRateRuleByVariable(const std::string& variable);
These functions return the designated object using the 'symbol' or
'variable' attribute. This is duplicate functionality as the
objects can be retrieved using getRule and getInitialAssignment.
However the other similar functions on the Model class retrieve
objects by the 'id' attribute. If future versions of SBML
introduce the 'id' attribute on Rules and InitialAssignments we
will need functions that act differently.
- 'comp' package-specific updates:
- The code that facilitates resolving files in relation to
ExternalModelDefinitions has been reviewed and made as robust
as possible.
- The ability to customize the way SIds and SIdRefs are
renamed during the flattening process has been added. A
class of type PrefixTransformer can be created and assigned.
This allows the user to add additional changes to the renaming
process. An example that illustrates the use of this facility
to also rename the 'name' attribute of objects originating in
subModels has been included in the examples/c++/comp directory.
- Unit tests have been added for the C API, and this exposed some
missing functions. The missing functions have now been added.
* Bug fixes
- There were issues with propagating the package-enabling
functionality through the children of an element. These issues
have been resolved.
- The converters that apply structural changes to the SBML document
used a setDocument function that expected a const SBMLDocument
object. Since the converters change the document, and in fact
immediately re-cast the document object, this function was both
confusing and inaccurate. There is now a setDocument function
that does not expect a const SBMLDocument object.
- There were some inconsistencies in the behaviour of LocalParameter
objects from SBML Level 3 in relation to the 'constant'
attribute. These have been fixed.
- 'layout' package-specific bug fixes:
- There were some discrepancies between the naming of arguments
to some constructors and the actual values being set using
these arguments. These have been corrected without changing
the resulting behaviour.
- Sometimes when an SBML Level 2 document with 'layout' and
'render' annotations was upgraded to SBML Level 3, the 'render'
namespace went missed from the annotation in the Level 3
document. This has been fixed.
- The 'render' package "objectRole" attribute was not being
correctly handled by the 'layout' code when the 'render' package
was disabled. This has been corrected.
- 'qual' package-specific bug fixes:
- Certain validation errors were not being correctly reported,
with either missing information or bad line numbers. This
problem has been fixed. Thanks to Nicolas Rodriguez for
reporting the problem.
* Configuration/build system changes
- WARNING: There is a bug in the Microsoft Visual Studio 2013
compiler that can cause strange output when doing repetitive file
reading and writing. We became aware of this when running our
unit tests on the comp flattening routines.
http://connect.microsoft.com/VisualStudio/feedback/details/
808199/visual-c-2013-floating-point-broken-on-first-use We
included a fix in our test files (libsbml/src/sbml/packages
/comp/util/test/TestRunner.c) but users should be aware that this
is an issue that may arise with this particular version of the
MSVC compiler.
- The CMake files have been updated to allow for the use when
libSBML is merely being referenced within a CMake build, rather
being specifically built.
- Both the GNU make and CMake configuration systems generate certain
header files that are specific to the libSBML build being
requested. Previously, libSBML source code included project
definition files for MSVC. These files were checked into SVN and
included within the released src archive distribution. Their
inclusion could potentially cause problems for people who wanted
to create different build configurations from the same source
code. Given that with CMake it is possible to generate MSVC
project files directly, we no longer include them in the source
archive distribution, and therefore, it is also not necessary to
include the header files generated by the configuration process.
Thus, these header files have been removed, and for CMake users
they will be generated within the build directory to facilitate
different builds using the same src tree.
- There were issues with classes exported across the Windows dll
interface. These have been fixed, and we have added additional
tests to catch these issues. Thanks to Andy Somogyi for reporting
the problem.
* Miscellaneous
- Python installers for various Linux distributions are now
available from the Open Build Service and via pypi. Links to the
relevant release can be found in the documentation on SourceForge
download pages for libSBML.
- The copyright statement for all example programs that we produced
ourselves has been changed to use a less restrictive license, so
that users can use the examples in their software without
limitation. The new license is the MIT license.
- The LICENSE.* files for libSBML now include a previously-missing
3rd party software acknowledgement for the Python argparse library
by Steven J. Bethard. (The argparse.py module is included with
libSBML distributions.)
- A number of extraneous files have been removed from the
documentation source directory (docs/src/); in addition, the
directory has undergone some minor restructuring.
- The Python API documentation lacked all the class constructors,
due to a bug in the libSBML documentation production pipeline.
This is now fixed.
EXPERIMENTAL RELEASE
* New features
- Support for the 'arrays' package according to the specification
dated September 11, 2013 (0.1 alpha) has been added to the
experimental release files. Note the extended MathML is accessible
via the ASTNode class; however, the current version version does
not yet include support for parsing to and from infix notation.
- 'multi' package-specific updates:
- The ability to extend ASTNodes and hence use extended MathML
has been added and thus it is now possible to use the attributes
on the elements of math as specified by the 'multi'
package. The example relating to this has been updated accordingly.
* Bug fixes
- 'render' package-specific bug fixes:
- On occasion when an L2 document with layout and render
annotations was being upgraded to L3 the render namespace
was missed from the annotation in the L3 document.
====================================================================
5.9.2-experimental (2014-01-31)
====================================================================
********************************************************************
*** Experimental only release ***
This is an experimental only release, that is, a release of those
files provided in the experimental folder of a full release.
This includes updates to all the SBML L3 packages that are not yet
accepted and undergoing development.
NOTE: An experimental release is not as rigorously tested as a
full release. Any changes/fixes listed as either referring to
stable or experimental code may change before the next full release.
********************************************************************
STABLE RELEASE
* New features
- The error messages relating to invalid SId syntax have been
improved to report the attribute causing the problem.
- The SBMLLevelVersionConverter has an option to ignore
unrequired L3 packages. However there was no way to access
this option when using the helper 'setLevelAndVersion'
function. This function has no been given a default argument
to allow users to specify the ignore packages option.
- A new version of the SBMLTransforms::evaluateASTNode function
is available; allowing the user to pass a map of ids and values
to the function rather than needing a whole model.
- Two new functions are available to allow users to determine
the compiled dependencies for the given instance of libSBML.
int isLibSBMLCompiledWith(const char* option) will take the name
of a dependency library e.g. 'expat' as argument and return a
positive integer this library was used to compile libSBML, 0
otherwise.
const char* getLibSBMLDependencyVersionOf(const char* option)
takes the name of the library and returns the version number
of this library.
* Bug fixes
- Some of the perl examples had syntax errors - these have been fixed.
Thanks to Joern Behre for reporting them.
- Some subtle bugs in the ASTNode class have been fixed. These
include returning 'bad allocation' as the name of a node;
failing to replace the definitionURL of a csymbol node;
not removing the units member when a node represents a function
and incorrectly reporting failure when setting user data to NULL.
Many additional tests have been added.
- The Borland compiler does not handle implicit casts and
unfortunately there was one such cast in the new code introduced
for 5.9.0. This has now been fixed. Thanks to Totte Karlsson for
reporting it.
- A memory leak in SBasePluginCreator has been plugged. Thanks to
Andy Somogyi for reporting it.
- The instances of the extensions loaded into the ExtensionRegistry
to facilitate plugins were not being correctly deleted. This
has been corrected.
- 'fbc' package-specific updates:
- The CobraToFbcConverter was losing the charge and the
capitalization of the formula when parsing the notes element.
This has been fixed.
* Configuration/build system changes
- Any necessary fixes to facilitate the use of the new version
of Mac OSX - Mac OSX 10.9 Mavericks - have been made.
EXPERIMENTAL RELEASE
* New features
- 'req' package-specific updates:
- The attributes in the req package are mandatory on all
elements. This caused huge numbers of errors when reading
files which used the req package but did not actually
need to attributes on many elements. Thus the error has been
suppressed.
- 'spatial' package-specific updates:
- The API for accessing compressed data has been improved
to allow access to this data using language bindings.
ImageData is now uncompressed during the first call to
this and saved for subsequent access. Previously the data
was uncompressed each time it was accessed.
- Functions to determine whether an attribute has been set;
i.e. the isSetFoo functions, have been provided where these
were missing. Boolean attributes that have not been set are
no longer written out.
* Bug fixes
- 'req' package-specific updates:
- The csharp binding for this package was not complete.
This has been corrected.
- 'spatial' package-specific updates:
- Errors relating to missing attributes have been sorted
to correctly relate to attributes required by the specification.
====================================================================
Version 5.9.0 (2013-11-14)
====================================================================
STABLE RELEASE
* Qualitative Models
- The first official version of the specification of the Qualitative
Models package is now available and thus the libSBML code for
'qual' is now included in the stable release of libSBML. This
means that all prebuilt binaries for the stable release will
include the 'qual' package code. The src archive containing
libSBML core code will continue to be available but there will be
additional src archives available in the stable branch: one
containing libSBML core plus all stable packages and additional
archives of the individual package code.
NOTE: The libSBML GNU make-based build system has not been updated
for packages. Thus, to build from src, it is necessary to
use the CMake build system.
* Layout
- The layout specification for SBML Level 3 has been accepted and
thus layout is now considered an official L3 package. Since
layout code has been included with libSBML for many years, this
has no impact on the code included within the stable release.
* New features
- There are two new function SBML_formulaToL3String(ASTNode_t*) and
SBML_formulaToL3StringWithSettings(ASTNode_t* L3ParserSettings_t*)
that provide reciprocal AST to infix functionality based on the
previously released L3Parser code.
- Within an SBML file it may be possible to infer units for
parameters based on their use within mathematical expressions. A
limited version of this functionality is now available using the
SBMLInferUnitsConverter which will return a document with units
attached to any parameters where it is possible to infer the
units.
- The SyntaxChecker class has a new function named
isValidXMLanyURI(std::string) that can check whether a given
string has the valid syntax for an XML anyURI datatype
http://www.w3.org/TR/xmlschema-2/#anyURI.
- There is a new base class ElementFilter that can be passed as an
optional argument to the getAllElements function. When
implemented this allows the elements to be filtered in whatever
fashion before being added to the List returned by the
getAllElements() function. The example getAllElementsWithNotes
illustrates this functionality.
- There is new functionality to facilitate replacing the id
attribute and all references to that id within a model. There is a
new base class IdentifierTransformer that when passed to the
function renameAllIds on the Model object will perform the
transformation specified by the IdentifierTransformer class on all
id attributes within the model. Users should derive their own
IdentifierTransformer class to perform the transformation they
require. The example setIdFromName illustrates this
functionality.
- The class IdList which allows users to create and manipulate lists
of 'id' strings has been made publically available. Previously
this class has been used internally by validation and conversion
code. It is now available in the sbml/util directory.
- There is a new converter, the SBMLIdConverter, that will allow a
user to replace several ids within a model. The converter requires
a comma seperated list of existing ids and another of the
replacement ids and will perform the necessary replacements.
- There is a new converter, the SBMLLocalParameterConverter, that
will allow a user to replace any localParameters with global
parameters. The promoteParameters example illustrates the use of
this converter.
- Following the results of the user survey the L3Parser has changed
the precedence of the '^' operator to be greater than a unary '-'
operator and the precedence of the '!' operator to be higher than
algebraic operators (and equal to the unary '-' operator). The
precedence of '&&' remains the same as the precedence of '||'.
- It is now possible to override the severity of errors logged or to
disable error logging completely. Any override is applied to the
errorLog until such time as the status is reset. This facilitates
the use of functions, that may on occasion log errors, in a
situiation were error logging is not required.
- There is a new function XMLErrorLog::changeErrorSeverity
(XMLErrorSeverity_t , XMLErrorSeverity_t, std::string package)
that will change the severity of errors within the errorLog based
on the package name supplied.
- The XMLErrorLog and SBMLErrorLog classes now have explicit
implementations for the copy and assignment constructors.
- The SBase functions getURI() and getPrefix() that return
information relating to the namespace of the object have been
made public. Formerly these functions were protected.
- The function removeTopLevelAnnotationElement(const std::string
elementName, const std::string elementURI = "") allows users to
remove a single top level element from the annotation node. If,
having removed the specified annotation, the overall annotation
node had no child elements the entire node was being deleted.
There is now an additional optional argument to this function,
bool removeEmpty, that allows the user to specify that they do not
wish the annotation to be deleted.
- There are two new function on the SBase class that allow an
element to be read from and written to an XMLNode. The XMLNode is
the underlying class used by the XML parsing layer of libSBML.
The function XMLNode* SBase::toXMLNode() returns a pointer to an
XMLNode that represents the SBase object; which will facilitate
the ability to look up attributes of an object by name. The
function void SBase::read(XMLNode& node,
XMLErrorSeverityOverride_t) invokes the read functionality on a
given XMLNode. This facilitates reading blocks of annotation that
may represent SBML in some way. It should be noted that these
functions are computational expensive in that they serialise to a
string. This drawback will be addressed in future libSBML
architecture but cannot be overcome with the current libSBML-5
implementation.
- All set or add functions in libSBML return a value from the
enumeration OperationReturnValue_t. There is now a function
OperationReturnValue_toString that will return a string
representation of the reason for the failure.
- 'layout' package-specific updates:
- The example code for layout has been moved to subdirectories
named 'layout' beneath the appropriate language subdirectories
in the examples directory. This reflects the structure used for
example code from all packages.
- The code reading and writing layout as an annotation has been
adapted to use the new read/write to/from XMLNode functions.
This greatly reduces the complexity and repetition of code
previously needed to perform these functions.
- 'comp' package-specific updates:
- All the validation rules have been fully implemented, up through
release 3 of the specification.
- Validation is now incorporated into the flattening process. If
there are no errors (or only warnings), the flattening process
will proceed to produce a flattened version of the model. If
there are errors, the flattening process will halt and return an
unchanged version of the model, with the errors noted in the
model's error log. This provides the user with detailed
information regarding why a document has failed to flatten as
well as detailing any warnings related to the document returned
on success.
- There is an additional option for the CompFlatteningConverter to
allow the users to request that the document not be validated
before or after flattening. This allows the user to produce a
flattened model irrespective of the validity of the original
model. Errors produced during flattening will still stop the
flattening process, as they make it impossible to flatten the
model.
- There is a further additional option for the
CompFlatteningConverter to allow the users to specify how the
required status of any unflattenable package should influence
whether flattening is attempted or not. By default this option
has a value of 'requiredOnly' which indicates that flattening
will abort if it encounters an unflattenable package that has a
'required' attribute of 'true'. Other options are 'all' or
'none'.
- The flattening routine has been extended to include all packages
available within the stable release: layout, qual and fbc.
NOTE: This means that the behavior of the flattening converter
will be different than it used to be: in the past, the converter
would either refuse to flatten a model with information from any
of these packages, or would leave unmodified package information
from the main model only. Now, all information from all
submodels will be properly integrated into the flattened model.
* Bug fixes
- The converter that expanded functionDefinitions was failing to
correctly deal with any csymbols it encountered in the math. This
has been fixed.
- Conversion between SBML Levels was not correctly handling the
units attributes that appear on the Model object in an Level 3
model. This has been fixed. Thanks to Pratham Shah for the report.
- The UnitDefinition::printUnits function had a mismatch between
integer and double values for the exponent; causing incorrect
values to be used on occasion. This has been fixed. Thanks
to Pratham Shah for reporting this.
- An annotation on a package element is permitted to use the SBML
namespace. However, this situation was logging an error. This
has been corrected.
- A memory leak in the TranslateSBML function has been plugged.
Thanks to Neil Swainston for reporting the issue and testing the
fix.
- 'comp' package-specific bug fixes:
- It is possible to construct a model which tries to delete an
object more than once. This was not being detected correctly.
This has been addressed and flattening code will report this
issue as an error.
- Similarly replacing both an object and independently replacing
a child of that object could also cause problems. These have
been sorted.
* Miscellaneous
- The API manuals for Java, Python and C# now include information
about the SBML Level 3 package extensions. In addition, many
content errors have been fixed, many formatting errors have been
resolved, new sections have been added, and finally, the overall
look and feel of the documentation has been updated (and hopefully
improved).
- The copyrights and license statements in LICENSE.txt/LICENSE.html
have been updated to reflect the inclusion of code from Ralph
Gauges (EML, Heidelberg, Germany), a small program from Sun
Microsystems used in generating documentation, and the PrettyTable
and Google Code Prettify utilities mentioned below.
- To help produce better documentation, the libSBML sources now use
(and incorporate) two utilities, PrettyTable 0.7.2 (in
src/bindings/python/doc-converter) and Google Code Prettify (in
docs/src). Both are redistributable under terms compatible with
the LGPL terms used for libSBML as a whole. Please see the file
LICENSE.txt or LICENSE.html for specific information about the
authors and license terms of these packages.
- A new, simple tutorial example of creating a model using the
Python language interface is now available in the Python
documentation. The example was kindly provided by Martins Mednis
(in 2012!).
EXPERIMENTAL RELEASE
* New features
* Distributions
- There is now a stabilized version of the specification for the
Distributions package and the experimental release now includes
code for this 'distrib' package. As is the case with all packages
included in the Experimental Release the binaries are built with
'distrib' enabled and the libSBML source archive includes the
source code.
* Multistate, Multicomponent and Multicompartment Species
- There is now a stabilized version of the specification for the
Multistate, Multicomponent and Multicompartment Species Package
and the experimental release now includes code for this 'multi'
package. As is the case with all packages included in the
Experimental Release the binaries are built with 'multi' enabled
and the libSBML source archive includes the source code. It
should be noted that it is not yet possible to extend the math
element; which the 'multi' proposal does; however the
definitionURL attribute can be used on the element to convey
the necessary information until libSBML provides the ability to
extend the math.
* Groups
- The 'groups' package code has been updated to include revisions
made to the latest specification.
====================================================================
Version 5.8.0 (2013-02-28)
====================================================================
STABLE RELEASE
* Flux Balance Constraints
- The first official version of the specification of Flux Balance
Constraints is now available and thus the libSBML code for
'fbc' is now included in the stable release of libSBML. This
means that all prebuilt binaries for the stable release will
include the 'fbc' package code. The src archive containing
libSBML core code will continue to be available but there will be
additional src archives available in the stable branch: one
containing libSBML core plus all stable packages and additional
archives of the individual package code.
NOTE: The libSBML GNU make-based build system has not been updated
for packages. Thus, to build from src, it is necessary to
use the CMake build system.
* New features
- The converter that expands SBML FunctionDefinition constructs has
been extended with an additional option that allows the user to
specify particular objects that should not be
expanded.
- We have changed the behavior of the converter that converts
between core Levels and Versions of SBML. This converter has a
boolean option, 'strict', that determines whether an invalid model
can be converted or produced. The code originally assumed that
'strict = false' refers to syntactic validity only, and so even
when the flag was set to false, the converter could fail to
convert a model if semantic validity was not preserved. This
interpretation has been changed because people felt it was not
useful; now, setting 'strict' to 'false' will always convert a
model, and if there are errors related to altered semantics, the
errors will be logged in the usual error log.
- Documentation for the 'comp' and 'fbc' C++ API is now available.
- 'comp' package updates:
- The C API for comp is now available with an example of
creating a model using this API.
- The majority of 'comp' package validation rules have now been
implemented and will be reported. Validation has also been
expanded to apply validation from core and other packages
(where appropriate) to any ModelDefinitions contained within a
comp model.
* Bug fixes
- LibSBML was still writing out the deprecated, and in some cases
removed, attribute 'charge' on a . This is no longer the
case.
- Some files caused the SBML Level and Version converter to enter an
endless loop. This has been fixed. Thanks to the good folks at
Persistent Systems Ltd. for reporting this.
- Reading an attribute with a value of type 'double' was failing
to report when a different type was encountered for the value.
This has now been sorted and will be correctly logged as an
error by the readSBML functions.
- The Level and Version converter was in some cases running unit
validation even if the user had specified that they did not
require strict unit validation. This was in fact done by libSBML
to allow an error to be logged; however, it has become clear that
the large computational overhead of running unit validation is
unecessary to merely log an error. Thus, if the user has declared
that they do not require unit validity, the checks will no longer
be applied.
- 'layout' package-specific bug fixes:
- The extension namespaces were not being correctly set. This
has been fixed.
- 'comp' package-specific bug fixes:
- The function 'renameSIdRefs' now functions properly for
SBaseRef-derived classes.
* Configuration/build system changes
- 'fbc' package-specific build system changes:
+ The GNUMake build system has not been implemented for
packages. Thus, to build from src, it is necessary to use CMake.
* Miscellaneous
- The structure of the API documentation has been slightly
overhauled to provide clearer access to relevant information.
EXPERIMENTAL RELEASE
* New features
- 'qual' package-specific new features:
- New functions getInputBySpecies(std::string) and
getOutputBySpecies(std::string) have been added to the
Tranisition class. These will return the first Input/Output
in the relevant list that matches the qualitativeSpecies
attribute value supplied as argument.
* Bug fixes
- 'qual' package-specific bug fixes:
- As validation of packages is starting to be introduced the
situation whereby a ListOfFunctionTerms can contain no
FunctionTerm elements but only a DefaultTerm caused issue
with the code that reports an empty ListOf object. This
has been fixed.
- 'render' package-specific bug fixes:
- Conversion between L3 and L2 annotation was not working as
expected. This has been fixed.
- The interaction between layout GraphicalObjects and the
render 'objectRole' attribute was removing the attribute
on occasion. This has been fixed and more fully tested.
====================================================================
Version 5.7.0 (2013-01-14)
====================================================================
STABLE RELEASE
* Hierarchical Model Composition
- The first official version of the specification of Hierarchical
Model Composition is now available and thus the libSBML code for
'comp' is now included in the stable release of libSBML. This
means that all prebuilt binaries for the stable release will
include the 'comp' package code. The src archive containing
libSBML core code will continue to be available but there will be
additional src archives available in the stable branch: one
containing libSBML core plus all stable packages and additional
archives of the individual package code.
NOTE: The libSBML GNU make-based build system has not been updated
for packages. Thus, to build from src, it is necessary to
use the CMake build system.
Known limitations of the 'comp' code:
+ Documentation has not yet been integrated into the libSBML
documentation.
+ Whilst the majority of validation rules have been implemented,
some will report the schema error id rather than the specific
comp validation rule id. The more involved rules involving
circular dependencies and in-depth referencing have not been
implemented. Validation of ModelDefinitions is implemented
with respect to comp validation, but not with respect to core
validation.
+ There is no C API for interacting directly with comp
objects. However the C API for reading and validation will
apply to the 'comp' package and flattening can be accessed via
the converter C API to call the CompFlatteningConverter.
* New features
- Convenience functions have been added to the SBMLExtensionRegistry
class to facilitate enabling/disabling packages using just the
package name. For example: disablePackage(const std::string&
package); enablePackage(const std::string& package);
- Previously the setNotes function accepting a string as argument
expected the string to contain the appropriate XHTML tags and
would not set the value of the element to a string that
did not meet this requirement. The functionality of the
setNotes(const std::string&) function has been extended to allow a
user to add an optional boolean argument addXHTMLMarkup. When the
value of this argument is 'true' the function will accept a
string; wrap it with the XHTML tag and set the element
to this.
- A C API for use with the converters has been added. Examples of
using the converter API are available in all supported languages.
- 'comp' package
- Access to the flattening routines has been restricted to the
CompFlatteningConverter. This allows options to be specified by
the user regarding ignoring other packages and whether to
produce a document with no comp namespace or a document with the
comp namespace but the required attribute set to 'false'.
- Validation has been implemented (see note above). This is called
via the checkConsistency function within core libSBML; which now
accesses a checkConsistency function in the comp plugin.
There are two categories of validation:
LIBSBML_CAT_GENERAL_CONSISTENCY and LIBSBML_CAT_IDENTIFIER_CONSISTENCY
which are enables or disable via the setConsistencyChecks()
function on the SBMLDocument (in libSBML core).
- The API for resolving external models has been updated to use a
registry of resolvers. This facilitates the addition of user
specific resolvers.
* Bug fixes
- The number of arguments expected by a piecewise function
was incorrectly assumed to be two. However, a piecewise
function can have a single argument. This has been corrected.
- Validation of SBML Level 3 was issuing a general warning about
missing required attributes but failing to identify which
attribute had caused the error to be reported. This has been
corrected. Thanks to Nicolas Rodriguez for pointing this out.
- Validation was failing to report the incorrect use of a function
call if the lambda function usesd expected no arguments. This
has been fixed.
- The L3Parser was not dealing with the sqrt function correctly.
This has been sorted.
- 'comp' package-specific bug fixes:
+ SBOTerms were not being correctly dealt with when they
occured on a Port element. This has been corrected. Thanks
to Chris Myers for reporting this.
* Configuration/build system changes
- Building the MATLAB binding for libSBML can be problematic due to
the need to identify and use the MATLAB mex compiler. When MATLAB
is being accessed over a network this can result in misleading
error messages. The build system has been improved to halt
compilation if the mex compiler is not available and to give a
clearer error message as to how to proceed.
- The build system has been updated to facilitate use with the
Mac OS X 10.8 operating system (Mountain Lion).
- 'comp' package-specific build system changes:
+ The GNUMake build system has not been implemented for
packages. Thus, to build from src, it is necessary to use CMake.
* Miscellaneous
- It was reported that gcc 4.7 found the various definitions of
isnan used within some parts of libSBML to be problematic. This
problem has been addressed and there are now functions util_isNaN
and util_isFinite in the util file that are used throughout
libSBML. Thanks to Totte Karlsson for reporting this.
- Where test files compare double values the comparisons have
been unified to use the util_isEqual(double, double) function
and thus avoid inaccurate results that can be machine dependent.
- This NEWS.txt file has reduced in size with NEWS dating from
prior to libSBML-5 now saved as an OLD_NEWS.txt file.
EXPERIMENTAL RELEASE
* New features
- 'Qual' package:
- The package specification redefined whether some attributes or
elements were required; the code has been updated appropriately.
- 'Fbc' package:
- The package specification added some optional attributes
and deprecated the use of strict less than or greater than;
the code has been updated appropriately.
* Bug fixes
- 'Render' package:
- In some cases the code was failing to add expected attributes
to particular elements. This has been corrected.
====================================================================
Version 5.6.0 (2012-08-03)
====================================================================
A note on the use of SWIG:
All aspects of libSBML-5.6.0 requiring SWIG are known to work with
SWIG version 2.0.4. Issues with later versions have been found
with respect to some areas of a libSBML build. These will be
addressed in future releases but for libSBML-5.6.0 we recommend
the use of SWIG 2.0.4.
STABLE RELEASE
* New features
- LibSBML now includes a binding to R (The R Project for Statistical
Computing). The binding is generated using SWIG and thus can be
built by users. Alternatively, the libSBML-5.6.0 release provides
R packages that can be installed directly into R.
- During HARMONY 2012, people discussed the issue of providing
support for Level 3 Layout in other SBML Level 3 packages. The
conclusion was that additional GeneralGlyph elements should be
added in order to support references to as-yet unknown types of
objects to be defined in Level 3 packages. These additional
elements have been included in the libSBML layout code.
- Although libSBML GNU makefiles have technically supported Python 3
for some time, there had been no feedback regarding the use of
Python 3 until recently. In response to the feedback, we have the
updated the build system to facilitate the use of different
versions of Python 3. On Windows, the libSBML release also
includes prebuilt installers for Python 3.2.
- The libSBML 'configure' command now accepts the additional option
--with-python-interpreter, which allows you to specify the Python
interpreter that should be used. This is useful when a system has
multiple versions of Python installed in the same directories, and
the version that should be used can be selected by explicitly
invoking the desired version. The option takes a full path, as in
"--with-python-interpreter=/usr/bin/python2.7".
- In order to facilitate the use of identifiers.org URLs, we added
an additional function, getSBOTermAsURL, to the SBase class. This
function returns a string representation of the SBOTerm in its
resolvable identifiers.org URL format; e.g.,
"http://identifiers.org/biomodels.sbo/SBO:NNNNNNN"
- The use of libSBML get and set functions in Python; a by-product
of the use of SWIG; is not particularly intuitive. LibSBML-5.6.0
applies a patch that allows the use of indexing in the python
environment i.e. model->getSpecies(0)->getId() can be replaced
with model.species[0].id. This patch is limited to Python 2 versions
at present. It does not adversely effact existing functions for
either Python 2 or Python 3. Many thanks to Gordon Ball
(Computational Medicine, Karolinska Institute) for creating and
sharing this patch.
* Bug fixes
- When using a copy of libSBML compiled with Level 3 package
support, attributes from the "sbml" namespace were not given a
namespace prefix when they should have been. This has been fixed.
- The L3Parser had some subtle bugs on some operating systems when
dealing with very long numbers or exponentials. These have been
tracked down and fixed.
- Although libSBML explicitly prevented History or CVTerm objects
from being added to the annotation of an object when that object
had no metaid attribute value, it was still possible to add
History and CVTerm directly to the annotation. This could lead to
invalid MIRIAM annotations because they lacked a metaid attribute
value. LibSBML now checks that an object to which any MIRIAM
compliant annotation is being added has a metaid.
- When returning a MIRIAM-compliant annotation, all CVTerms were
lost if just one had an invalid qualifier. This has been fixed.
- The decompression code uses a macro OF that is not defined for all
versions of linux. A definition has been added to avoid problems.
Thanks to Kai Wohlfahrt for reporting this.
- There was a potential memory leak issue while registering
SBMLConverters. Thanks to Akira Funahashi and his team for spotting
the problem and supplying a fix.
* Configuration/build system changes
- As SBML Level 3 packages and their code become available, there is
a increasing number of possibilities in the matrix of what a
particular instance of libSBML has included. In order to
facilitate code that may rely on the presence of one or more
packages, the CMake configuration process has been improved to
include a set of flags that identify the packages included in a
particular build. The file convertLayout.cpp in the examples/c++
directory illustrates the use of these #define's.
- The MATLAB bindings failed to build on Mac OS X whenever an Xcode
version was used that did not include the Mac OS 10.6 SDK. This
has been fixed.
- The CMake build system has been made more resilient to finding
include files from previous libSBML installations.
EXPERIMENTAL RELEASE
* New features
- The Qualitative Modeling package "qual" is now supported.
- The R binding is available for all packages.
- 'Groups' package:
- At HARMONY it was agreed that the 'group' element needed an
additional attribute 'kind'. This has been added.
- 'Spatial' package:
- There was an issue with the encoding of image data for the
spatial package. This has been solved by providing a function
that will uncompress the data and return it to the user in a
usable form.
- 'Comp' package:
- Instantiated Model objects in Submodels now are set to have that
Submodel as their parent. The SBMLDocument is still set to the
document of the original Model, so that its own Submodel
references can be properly tracked down.
====================================================================
Version 5.5.0 (2012-05-11)
====================================================================
STABLE RELEASE
* Revision of annotation and SBML MIRIAM RDF
Following discussions on the libsbml-development mailing list
we have adjusted the way functions affecting annotations interact
with the SBML MIRIAM RDF classes for History and CVTerms.
- Previously any annotation was assessed and if it contained
MIRIAM RDF the corresponding classes were created and the original
annotation XMLNodes relating to these objects was removed from the
annotation. When the annotation was retrieved the xml was reconstructed
from the History and CVTerm classes. This is no longer the case.
In 5.5 we have changed the code so that the MIRIAM objects are
created BUT the original annotation remains intact. It is only
reconstructed if the code has directly altered the History or CVTerm
classes via the class API, otherwise the original annotation
is returned.
- We reassessed the use of the appendAnnotation function. Previously
any annotation was appended regardless of whether the existing
annotation already contained a top level element with the same
namespace. We considered this to be a bug, since it would result
in invalid SBML. Thus, the appendAnnotation function now checks
whether there is an existing top-level element with the same
namespace of any it is to append and does not append if it
encounters a duplicate. It returns an OperationReturnValue of
LIBSBML_DUPLICATE_ANNOTATION_NS to indicate this to the user.
- In our reassessment of appendAnnotation we came to the conclusion
that it was virtually impossible to correctly determine the intent
of the user in cases where both the existing and appending annotation
contained RDF that complied with the SBML MIRIAM RDF History or
CVTerms. Thus the appendAnnotation function no longer treats RDF
as a "special" case. If there is existing RDF, no further RDF will
be appended. Users wishing to alter existing History or CVTerms should
use the API provided or construct their own annotation and use the
setAnnotation function.
- New functions removeTopLevelAnnotationElement(std::string name) and
replaceTopLevelAnnotationElement(XMLNode* annotation) have been
introduced to allow users to remove or replace a single top level
annotation element whilst conserving any other annotation that may be
present.
* New features
- There is now a version of the infix to MathML parser that is independent
of L1 constructs. This parser is available as three functions:
SBML_parseL3Formula(""), SBML_parseL3FormulaWithModel("", Model*) and
SBML_parseL3FormulaWithSettings("", L3ParserSettings *). There is
a new object L3ParserSettings that allows the user to tailor the
results of the function, for example to parse "log" as natural log
or log base 10; to allow users to enter units on numbers as "10m" or
to disable this feature along with some other options.
- The getLine() and getColumn() functions did on some occasions
return a maximum int value when the correct lines and columns
could not be determined. This produced inconsistent behaviour as
some XML parsers return '0' as if they cannot determine location.
This has been unified and libSBML's getLine()/getColumn() will now
return '0' if the location is unknown.
- There is a new class SBMLDocumentPluginNotRequired which allows a
package code to derive from an object where the required attribute is
considered 'false' until it has been set. Note this is not a "default"
value it is just the value with which the boolean attribute is
initialised. The existing SBMLDocumentPlugin class assumes a value
of 'true'. This is in fact a change in behaviour but since users
should set this value it should not cause a problem.
* Bug fixes
- Tracker issue#3509843
The annotation element in SBML should contain only elements. Schema
validation reports this but libSBML was not. This has been corrected.
Thanks to Camille Laibe for still running schema validation and reporting
the issue.
- There was a subtle bug involving accessing a NULL ModelHistory object.
This has been fixed.
- Area units were not being properly converted between SBML Levels. They
are now.
- An AST plus or times node that was created with more than two children
was written out as only having two children; thus losing information.
This behaviour has been corrected.
- Functions used to add objects to other objects in libSBML check that
the SBML Level/Version/Namespaces of the objects correspond. However
these checks were imposing the condition that the number of namespaces
on both objects be the same. This is not a necessary condition and
the code has been relaxed. Thanks to Chris Myers for pointing it
out.
- In XML if the sbml element is not prefixed but the only declaration
of the sbml namespace is, then technically the sbml element has not
been placed into the sbml namespace. LibSBML was failing to report this.
This has now been fixed.
- There was a curious indentation issue that arose if notes used text as
child elements of an element that also had another non-text child element.
This could proliferate and cause models where the indentation ran into
the thousands of spaces. This has been sorted out and no longer
happens.
* Miscellaneous
- The uninstallation instructions were out of date. This is now
updated, for example to mention that the Mac OS X installer does
in fact provide an uninstall script.
* Configuration/build system changes
- Doxygen versions 1.7 and 1.8 are now supported, but due to
differences in the new version, older versions of Doxygen are no
longer supported. The 'configure' script will now only accept
version 1.7 at minimum.
- Some changes have been made to facilitate building libSBML
with the Borland compiler. This included standardising the use
of #ifdefined rather than using #if. Thanks to Totte Karlsson for
the effort of getting this to work.
- The CMake build has been updated to facilitate the use of the Intel
compiler.
EXPERIMENTAL RELEASE
* New features
- New macros provided so that SBMLNamespaces objects may easily be created
to allow package elements to be extended by other packages (for example:
a comp:modelDefinition to have a layout:listOfLayouts). These macros
now used throughout package system when creating new package elements.
- In 'comp':
- ExternalModelDefinitions have a new function, 'getReferencedModel'
that tries to find the referenced document and model, if the reference
URI is a 'file' type. If the SBMLDocument has its fileURI set, this will
use that to search for relative filenames.
- With the above in place, automated flattening now works with
ExternalModelDefinitions that point to local files.
- May now use ModelDefinitions to create the main Model of an SBMLDocument.
======================================================================
Version 5.4.1 (2012-02-24)
======================================================================
STABLE RELEASE
* Bug fixes
- The MATLAB binding function isSBML_Model had a subtle bug
that meant it could crash in certain circumstances. This has
been addressed. Thanks to Ying Zhang for the report.
- When determining units from rate rules unit checking failed
to catch the case where time had been redefined as dimensionless.
This is now fixed. Thanks to Kieran Smallbone for reporting it.
- Documentation files for MATLAB were missing from the dmg
installers for MacOS. This has been fixed.
- 'getElementBySId' could potentially return initial assignments,
event assignments, assignment rules, and rate rules that assigned
*to* the SId in question, instead of returning the actual element
with the SId. This has been fixed.
EXPERIMENTAL RELEASE
* Bug fixes
- LibSBML will now automatically register any packages that are
enabled in code. Previously libSBML could be built including
package code but users needed to perform an additional step
to actually use it. This was causing some confusion.
======================================================================
Version 5.4.0 (2012-02-16)
======================================================================
STABLE RELEASE
* New features
- Basic validation of MIRIAM compliant RDF annotations that
are parsed into ModelHistory or CVTerm objects has been
added to facilitate the use of annotations that are not
yet quite complete. Warnings are now issued relating to
whether or not the rdf:about attribute accurately points to
the metaid of the containing object and whether the object
that was created was consisdered a valid object in the
attributes and subelements that are required.
- It has been noted that unless the checkConsistency function is
called immediately after reading an SBML document, it does not
report all potential errors. Since users may wish to apply the
function to a model being constructed this seems to be a
disadvantage of the function. However, as this function has been
in libSBML for many years it was decided not to alter the
functionality but to add a new function 'validateSBML' which, when
called from a document object, will force a write and read of the
in memory object and report validation errors for the object as
written. Thus, all errors will be reported with line numbers that
would correspond to the document if written out.
- Conversion of a model containing FunctionDefinitions to L1
is now supported.
- Conversion of a model containing InitialAssignments to L2V1
and L1 is now supported.
- There is a new isSetBody() function for the FunctionDefinition
class that returns true if the body of the lambda function
has been set.
- The MATLAB/Octave binding function OutputSBML now takes an
additional optional flag (extensions_allowed defaulting to true)
that allows a user to specify whether the MATLAB_SBML structure
should be considered valid if it contains any additional fields.
- Multiple namespaces using the same prefix are not allowed in XML.
The function that added XMLNamespaces in libSBML checks for an
existing namespace with the prefix of the namespace being added
and overwrites the original if a duplicate is found. This
behaviour meant that it was possible to unintentionally remove
the sbml namespace. The behaviour of the function has been altered
such that if the namespace being added has a prefix that duplicates
that of the existing sbml namepace then the function returns failure
and the new namespace is not added. Should the sbml namespace need
to be replaced it can be done by using the remove function before
adding the new namespace.
- SBase objects that occur in Level 3 package code need to keep
track of the namespace of the package to which they belong. However,
a conversion might alter the namespace; for example with layout
where the level 2 layout annotation has its own namespace.
The getURI function has been altered to adopt the approach used for
tracking the core sbml namespace of any object i.e. the namespaces of
the containing document take precedence and the namespace stored on
an actual object is only considered if it does not belong to a document.
An additional function getElementNamespace() can be used to obtain
the namespace stored on the actual object.
* Bug fixes
- Tracker issue #3472034
Line numbers were not being reported as accurately for L2 models
as for L3 models. We have overviewed all the code relating
to line numbers and improved the accuracy and consistency
of line numbers reported. Thanks to Raimund Rönn for the report.
- Tracker issue #3471902
The documentation for the readSBMLFromString function claimed it
prepend the xml encoding if it was not present. In fact, the
function did not do this. This has been corrected. Thanks to
Raimund Rönn for reporting it.
- Tracker issue #3466104
Fixed inaccuracies in the documentation. Thanks to Raimund Rönn
for his careful reading and taking the time to report the errors.
- Validation failed to notice that csymbol avogadro was not valid
in Level 2. This is fixed.
- Exceptions were not being correctly wrapped for language bindings
causing potential code crashes. This has been corrected.
- A number of minor issues with regard to conversion between levels
and versions have been fixed. These include conversion from L3 to
L1 failing to include a compartment; stopping conversion of models
with hasOnlySubstanceUnits = true to L1; correcting conversion of
L3 models with localParameters and preventing conversion of L3
models where Events had Priorities or where the csymbol avogadro
had been used.
- The internalConsistency validator was incorrectly reporting that
the an Event must have a listOfEventAssignments in a level 3 model.
This restriction was actually removed in SBML Level 3.
- The MathML subset used by SBML has plus/times/and/or/xor. In the
MathML specification all these functions are nary functions i.e.
they are permitted to have any number of arguments including zero.
LibSBML had in fact restricted these functions to two or more
arguments. This inconsistency has been corrected.
- Validation was failing to apply constraints to L3 LocalParameters.
This has been corrected.
- Using a value for a units attribute that was not a builtin unit or
the identifier of a UnitDefinition is invalid. However, the way in
which this was reported varied according to which unit attribute
caused the issue. This has been standardized to produce an id
warning about a dangling UnitSIdRef and (if selected) unit warnings
relating to the impact of the actual attribute.
- In certain situations libSBML would write out both the stoichiometry
attribute and a stoichiometryMath element. Since these are mutually
exclusive this should not happen. It has been fixed.
- The code that checked for consistent namespaces when adding one
object to another was too restrictive. This has been sorted. Thanks
to Chris Myers for reporting it.
- Conversion between levels was inconsistent when it came to dealing
with rational stoichiometry. This has been fixed.
- The tests for the Perl bindings use of BigInt caused ActivePerl
to crash on a Windows system. A workaround to prevent this has
been put in place.
- LibSBML class constructors currently throw exceptions rather than
create objects with invalid SBML namespace/level/version
information. This could potential result in code crashing
when using one of the language bindings. These have now been
rigorously tested and measures put in place to prevent such
crashes from happening.
- The SBMLTransforms class now correctly evaluates piecewise
expressions.
- The MathML infix converter now deals correctly with unary plus
and times.
- The help text for classes and methods in the Python library were
previously unformatted, marked-up Doxygen input text, making the
help text unhelpful. We wrote a new conversion program to process
the marked-up text and produce plain-text documentation strings in
the Python library. Users can now use the standard Python 'help'
function to obtain help as expected.
- The StripPackageConverter was not working as expected. This has been
fixed.
* Miscellaneous
* Configuration/build system changes
- As libSBML has evolved over the years the header files
included by a given code file have expanded. This makes it
necessary to recompile many files when only a single header
file has been altered. In an attempt to eliminate unnecessary
dependencies we have done an overview of the code. However,
merely replacing existing include files with those that are
actually required may cause build issues for other users. Thus we have
introduced a LIBSBML_USE_STRICT_INCLUDES flag. This flag is
not defined by default and thus no includes will change. Defining
the flag means that code will only include the minimum number of
necessary header files. We encourage people to consider using this
flag as it may become the default behaviour in later versions of
libSBML. The flag is available via the CMake build interface.
EXPERIMENTAL RELEASE
* New features
- Code for the 'render' extension is now available.
- All package code has been updated to use the new line/column
number reporting mechanism.
- Package language bindings have been updated to properly wrap
exceptions and use covariant return types.
- There is a new example with the 'layout' package that demonstrates
conversion between the layout and render L3 package representation
and the L2 annotation representation.
======================================================================
Version 5.3.0 (2011-12-23)
======================================================================
STABLE RELEASE
* New features
- Errors reported by the matlab binding function 'OutputSBML'
have been improved to provide information about why the
structure is invalid.
- There is a new virtual function on an SBase object 'setElementText'.
This enables a user to store text that the read functions encounter
within an XML element. Previously such text was ignored as no such
text existed as a part of an SBML Core element. Some L3 packages
may use this XML feature and can thus utilize this function.
- There is a new function on an ASTNode 'returnsBoolean' that returns
a predicate indicating that the math of the ASTNode will return a
boolean result or not. The function has an argument of type Model
with a default value of NULL. Where necessary, that is, when the
ASTNode involves a call to a user defined function, the check is
done in the context of the supplied argument Model or the parent
Model of the particular ASTNode if no argument is supplied.
- There are new functions to facilitate the use of packages. The
SBMLExtensionRegistry has a 'disableUnusedPackages' function which
disables registered packages that are not being used by the target
SBMLDocument. The base SBMLExtension class has a function 'isInUse'
that will establish whether a package is being used by the target
document.
* Bug fixes
- The conversion code incorrectly converted an L3 model where the
stoichiometry was changed by a RateRule. This has been fixed.
- Conversion from L3 to L2 was failing to write out the spatialDimensions
attribute of a Compartment regardless of the value. It now
correctly writes out a non default value.
- The RegisterConverters class was not included in the
LIBSBML_CPP_NAMESPACE. This has been corrected. Thanks to
Christoph Flamm for letting us know.
- There was a slight discrepancy in how the assignment operator
for the SBMLExtensionNamespaces was behaving depending on OS.
This has been sorted so that behaviour is the same for all.
- Tracker issue #3438777
The XercesParser was failing to find the line and column number
definitions with some compilers. This is now fixed. Thanks to
David Fange for reporting it.
- Tracker issue #3441615
There was an issue with SWIG, C# and MSVC compilers caused by
conflict with MS MFC and GDI classes. This has been fixed by
explicilty excluding these classes. Thanks to Totte Karlsson
for reporting this.
- Tracker issue #3461400
Writing out an L2 model with Layout annotation caused duplicate
annotations to be written out. This has been fixed. Thanks to
Raimund Rönn for reporting the problem.
- Tracker issue #3462055
A couple of typos have been corrected. Thanks to Raimund Rönn
for taking the time to report these.
- The documentation for Python did not omit internal libSBML
methods. This is now fixed, and the docs more properly reflect
the public libSBML API.
- We have fixed a number of errors in the Python API documentation,
such as return values being reported as "char" when in fact they
were strings, and have filled out more of the missing
documentation.
- We have fixed many other small documentation bugs and filled out
more of the missing documentation.
* Miscellaneous
- The matlab binding contains a function 'isSBML_Model' that checks
that the structure contains all the fields that are expected
in an SBML model. This function is used by OutputSBML to determine
if the model can be written out. Previously this function would
only report that a model was valid if each of the substructures had
the expected fields and no additional fields. This behaviour has
changed to allow users to include extra information within the
structure. Thus a model is considered invalid if it does not have
all the expected fields; additional fields are ignored.
- The libSBML examples code has been ported to all language bindings;
so all examples exist in each language.
- The API manuals for languages besides Java now feature a more
modern and attractive appearance.
* Configuration/build system changes
- Using CMake to link against the static runtime libraries on windows
when using libxml2 it was necessary to manually add links to other
libraries. The build system now takes care of these.
EXPERIMENTAL RELEASE
* New features
- There is now code to support the L3 render package.
======================================================================
Version 5.2.0 (2011-11-11)
======================================================================
* Download directory structure
- Prior to this release we have provided stable libSBML source
code and associated binary installers, but only provided the
source code of the L3 packages that are in beta development.
In response to requests to provide binaries including the package
code, we have slightly reorganised the directory structure of
the 5.2.0 release.
The top level directory remains as expected
https://sourceforge.net/projects/sbml/files/libsbml/5.2.0
This directory contains two sub-directories 'stable' and
'experimental'.
1) files/libsbml/5.2.0/stable
This directory contains the source code and various binary
installers arranged in a structure identical to that used
in the 5.1.0-b0 and 5.0.0 releases.
2) files/libsbml/5.2.0/experimental
This directory also contains two subdirectories.
a) files/libsbml/5.2.0/experimental/src
This directory contains the archives of the beta source
code for the L3 packages; similar to those previously contained
in the separate top level directory /5.1-packages-beta/.
b) files/libsbml/5.2.0/experimental/binaries
This directory contains binaries that include the stable
source code plus the beta code for ALL of the packages. It
should be noted that these binaries may be untested and
are provided 'as-is' to facilitate experimentation with L3 packages.
* New features
- There is a new method 'removeFromParentAndDelete()' that enables
the complete removal of child objects from any SBase object.
- There is a new method 'getAllElements()' that returns a List
containing all the children, both direct and indirect, of the
object on which it is invoked.
- There are new methods that will rename any references to
SId and UnitSId throughout the model.
- A new converter has been added that will convert the units
of a model to base SI units.
- A new converter has been added that will order AssignmentRules
and InitialAssignments in such a way that forward references
are avoided.
- The validation API has been refactored to facilitate the use of
third-party validators. The existing API has been maintained; in
addition, it is now possible to call any external validator,
regardless of programming language, using the new validator API.
- There is a new UserData object on the SBase class. This
facilitates storing additional application-specific information
that may be useful to particular software.
- A new function, appendFrom(ListOf *), on the ListOf class allows
the contents of one ListOf object to be appended to another.
- There are new functions getNumRegisteredPackages() and
getRegisteredPackageName(index) that allow the user to establish
which Level 3 packages have been enabled in the instance of the
libSBML library being used.
- The C API for interacting with the package plugin mechanism
has been expanded.
* Bug fixes
- The functions that add SBase objects were only checking that the
object being added had the correct Level and Version of SBML. It
failed to check that any SBML namespaces used also matched. This has
been corrected.
- Error checking failed to catch SBML L3 models with an empty
. Fixed.
- If the units of an object referred to a UnitDefinition that had no
, the warnings generated by the unit checking code
were inconsistent. These have now been standardized.
- The UnitDefinition::isVariantOfMass() function did not check that
the exponent of any unit was '1'. This has been fixed.
- The postal address for the Free Software Foundation was out of
date in all files where it was mentioned. Fixed.
- It was (erroneously) possible to allow the SBML namespace to be
replaced and lead to the production of a document with no SBML
namespace. Fixed.
- Some of the older libSBML C code called the exit() function when a
memory allocation error occurred. This is not an ideal way to
handle a problem. Since the functionality has existed in libSBML
for many years it has been removed from a standard build but can
be reenabled using a configuration flag should any users wish to
do so.
- The point at which an MSVC call to the c_str() function tidies up its
memory usage when running in debug mode is different to the point at
which the memory is tidied up when running in release mode or with other
C compilers. This means that a C function that returns a char * which
is created using c_str() will return an invalid pointer. Unit testing
does not highlight this issue as it only occurs in an MSVC Debug build.
Duplicating the string before returning it solves the problem.
LibSBML-5.1.0-b0 had three new functions where the string duplication
was missing. This has now been fixed.
- In the layout code, adding a to a produced
an element with the wrong name. This has been fixed; a fix that
revealed further issues with the constructors for packageNamespaces.
This issue has also been fixed. Thanks to Chris Myers for reporting
the original problem.
* Miscellaneous
- There were inconsistencies arising from the naming of functions
related to packages; some being named doPkgFoobar and others
doPackageFoobar. All of these functions now use the doPackageFoobar
naming format. Existing functions using the Pkg format will still
work but should be considered deprecated.
- The text of many SBML validation and consistency-checking error
messages have been edited to hopefully improve readability,
clarity and consistency.
- All example files have been overhauled and new ones added.
- Users reported that some of the SBML files in the "test-data"
directories were invalid, not realizing that this was actually
intentional because the files are used for testing libSBML. In
order to avoid confusion, the files containing invalid SBML have
been renamed to append '-invalid' to their file names.
- The files located in the validator/test-data directory, used for
testing the SBML validation code, have been reorganised into
smaller subdirectories. The new subdirectories group the files
according to the type of validation being tested.
* Configuration/build system changes
======================================================================
Version 5.1.0-b0 (2011-08-25)
======================================================================
This release includes a provisional implementation of support for an
SBML Level 3 package for "layout" based on code from Ralph Gauges of
the COPASI group. No other packages are included with this release;
this applies to the "groups" package as well, which was in the beta
release but is not in this release. However, the SVN repository on
SourceForge has a branch for packages where ongoing development is
proceeding, and as specifications and code for SBML Level 3 packages
are settled, future libSBML releases will include stable versions of
packages as they become available. Source code for beta versions of
packages will also be made available from a separate download area,
http://sf.net/projects/sbml/files/libsbml/5.1-packages-beta
* New features
- The inline formula code will now deal with a unary plus.
- There is a new API for performing conversions between SBML Levels,
Versions, and SBML Level 3 packages on an SBML Document. Existing
functions such as setLevelAndVersion() have been reimplemented
using the new API and so will continue to work as expected. The
new API will facilitate the introduction of conversion routines
that deal with Level 3 packages.
- The MATLAB_SBML Structure produced using the MATLAB/Octave bindings
now includes fields for the SBML Level and Version on all
sub-structures.
- There are new functions getElementBySId() and getElementByMetaId()
that will traverse the entire document and return a pointer to an
SBase object with the SId or MetaId supplied.
- There are new functions on the SBO class to reflect the new top-level
names introduced in the SBO ontology.
* Bug fixes
- There were a couple of places where the clone methods failed to
create a proper copy. These have been fixed. Thanks to Nicolas
Rodriguez for reporting this.
- The introduction of the ability to place a units attribute from
the SBML L3V1 core namespace into a MathML element meant that
the code for reading and writing MathML needed to be fully aware
of which SBML namespaces it was working with. This has been done.
- Tracker issue #3296228
Importing a file with a large number of validation errors into
MATLAB was causing memory issues within MATLAB. The handling of
errors has been improved to avoid this situation.
* Miscellaneous
- Necessary updates to support the latest version of swig.
* Configuration/build system changes
- The introduction of Mac OS X 10.7 Lion produced issues with both
XCode 4.0 and MATLAB. These have been addressed.
======================================================================
Version 5.0.0 (2011-04-14)
======================================================================
This is the first release of libSBML-5.0.0.
There are three types of potential changes in this release compared to
the beta 5.0.0 release: changes that relate to support for SBML Levels
1, 2 and Level 3 Core; changes that relate to support for L3 packages,
and changes that were introduced in libSBML-4.3.1. These are listed
separately below for convenience.
This release includes a provisional implementation of support for an
SBML Level 3 package for "layout" based on code from Ralph Gauges of
the COPASI group. No other packages are included with this release;
this applies to the "groups" package as well, which was in the beta
release but is not in this release. However, the SVN repository on
SourceForge has a branch for packages where ongoing development is
proceeding, and as specifications and code for SBML Level 3 packages
are settled, future libSBML releases will include stable versions of
packages as they become available. Source code for beta versions of
packages will also be made available from a separate download area,
http://sf.net/projects/sbml/files/libsbml/5.0-packages-beta
-------------------------
Known limitations
-------------------------
At present, libSBML 5 will not attempt to convert a model between SBML
Levels (e.g., from Level 3 to Level 2) if the model uses SBML Level 3
packages. This applies to SBML layout elements and annotations too.
Our aim is to support conversion between L2 layout annotations and the
L3 layout package, but this functionality will be introduced in a
future release of libSBML.
--------------------------------------
SBML Level 3 Package extension changes
--------------------------------------
* New features
- The extension mechanism for supporting SBML Level 3 packages now
includes a full C API.
- LibSBML now has greatly improved detection and error reporting of
unimplemented/unknown SBML Level 3 packages.
* Bug fixes
- The "layout" extension now provides more seamless access to L2
layout annotations in L2 models.
* Additional documentation
- There is now more detailed and integrated documentation describing
the use of CMake as an alternative to GNU make for building and
installing libSBML.
-------------------------------
General libsbml-core changes
-------------------------------
* New features
- There are now functions will set the Qualifier on a CVTerm from a
string; in addition to the existing functions that use values from
the enumeration of QualifierTypes.
- SBML L2 contains an number of attributes that have default values.
LibSBML does not write these values out. This behaviour has
changed slightly for libSBML-5. If a user explicitly sets the
value of an attribute to be the default value; then it will be
written out. For example setting the spatialDimensions of an L2
to be '3' will result in a document where this is
explicitly written.
- In SBML L3, it is possible to have a compartment with no
spatialDimensions attribute declared. LibSBML-4 would not convert
such a model to L2, since an L2 compartment has spatialDimensions
'3' by default. This behaviour has been altered for libSBML-5 and
if the user opts for nonstrict units when applying the conversion
function, a compartment with no spatialDimensions declared will be
converted.
- There are additional functions XMLNamespaces::getNumNamespaces and
XMLAttributes::getNumAttributes. These are aliases for the existing
getLength functions but are more in keeping with the rest of the API.
* Bug fixes
- The conversion functions for converting an L2 model to L3 were
failing to set the stoichiometry value. This has been fixed.
- The internalConsistencyCheck was failing to report when required
attributes or elements were missing from an object. This has been
corrected.
- The XMLConstructorException was not exposed. It is now.
* Configuration/build system changes
- The header files libsbml-config-win and libsbml-config-unix have
been replaced with a single libsbml-config-common file.
- The organization of the source directories has been radically
changed, moving the previous directories src/annotation, src/math,
etc., to become subdirectories of src/sbml. This allows us to
entirely avoid a step that was previously to create a copy of all
the header files into /include. That copy operation lead
to a lot of wasted time when debugging code and accidentally
making changes to the wrong file.
- The 'win' directory that previously containing MSVC project files
has been removed. The addition of packages made maintaining this
directory problematic. The CMake build system provides an easy
alternative for generating any project files required.
-------------------------------
Libsbml-core changes from 4.3.x
-------------------------------
Also available in Version 4.3.1 (Released 2011-03-29)
* New features
- There is now API documentation for the MATLAB and Octave bindings.
* Bug fixes
- Certain versions of the libxml2 parser fail to catch a prefix
being assigned to an empty namespace. Additional code has
been included to catch this error if the version of libxml2 has
not reported it.
- The improved build of the MATLAB interface proved unsuccessful
in cases where the user had a startup.m file that changed
directory. This has been corrected.
- The SyntaxChecker::hasExpectedXHTMLSyntax(xhtml) function had not
applied the relaxation on the XHTML restrictions for SBML
Level 3. This has been sorted out.
- The SyntaxChecker::isValidSBMLSId(string) function was incorrectly
reporting that an empty string was a valid id. This has been corrected.
* Miscellaneous changes
- Choosing --enable-universal-binary on MacOS 10.5 now includes all
three of the following architectures: i386, x86_64, ppc. It
previously only selected i386 and ppc.
Also available in Version 4.3.0 (Released 2011-03-04)
* New features
- There was some disagreement about what libSBML when asked to write
out a required attribute in Level 3 where the value of the attribute
had not in fact been set. Following discussion within the libSBML
team and subsequently the libsbml-development mailing list it was
decided that the appropriate behaviour was to ignore the attribute
when writing out the model and thus not inadvertently give the
impression it has a value.
- Conversion from Level 3 to Level 2 involving variable stoichiometry
has now been implemented and this feature of L3 can now be converted
to L2.
- There is a getSupportedNamespaces function on the SBMLNamepaces
class which returns a List of the SBMLNamespaces supported by
this version of libSBML.
- Tracker issue #2082682
The unit checking of math involving the root function has been
fully addressed. Previously unit checking was only applied to
the expression root(n, x) if n was an integer. It can now handle
any value of n including an expression.
- Checking of the units of measurement in math expressions has been
further improved to provide checking to a deeper level with the
overall math of the model. Thus, the units of the expression a^b
can now be validated even if the value of b is assigned b = c and
the model states elsewhere that c = 2.
- SBML Level 3 has no default attribute values and thus a number of
new isSetXYZ functions were introduced in libSBML-4.1.0. Users
pointed out it was inconsistent that these functions returned
'false' for a Level 2 model in which the attribute concerned *did*
have a default value. Thus, any newly introduced isSetXYZ
function that deals with an attribute that had a default value in
Level 2 will now return 'true' regardless of whether the user has
explicitly set the function or not.
- The MATLAB and Octave bindings now include the function
OutputSBML() for writing the structure used in these environments
back to SBML. This function has previously shipped as part of
SBMLToolbox.
- There is an additional constructor for an SBMLDocument object that
takes an instance of SBMLNamespaces as its argument. This is similar
to the constructors for all SBML objects.
- There is now a custom manifest file included in the Java JAR file.
It contains the version number of the libSBML release. An example
program demonstrating how to access the information is included
as the file examples/java/printLibSBMLVersion.java.
- An API manual for the C# language bindings is now available. It
is not as complete as the C++ and Java versions, but it is quite
close.
* Bug fixes
- The InternalConsistency validator was failing to report the absence
of required attributes or elements within an in-memory model.
This has been sorted out with a large number of tests added.
- Tracker issue #2960646
When writing out an SBML model containing the root function
Octave was failing to convert the formula into the appropriate
MathML. This has been fixed.
- The code that checks for allowed functions in a Level 1
document was reporting an error if it encountered one of the
predefined functions. This has been corrected.
- Checking for NULL references has been greatly improved
throughout.
- The appendNotes function was not correctly checking for the
xhtml namespaces declared at the top level. This has been
fixed.
- In some cases the facility to create new 'Bags' within CVTerms
was not performing as expected. This has been sorted out.
- Validation previously failed to report an error if a lambda
function contained only arguments and no actual function
body. This has been fixed.
- LibSBML checks that each element within a model being read has
the correct default namespace. However, if the namespace had been
'declared' via a prefix this check was not being performed.
This has been corrected.
- The code that checked that the top-level elements of an annotation
had unique namespaces was incorrectly comparing prefixes rather
than actual XML namespaces. This has been fixed. Thanks to
Waclaw Kusnierczyk for reporting this and supplying the fix.
- The XML namespace for a