relax releases

From relax wiki
Revision as of 19:06, 11 September 2014 by Bugman (talk | contribs) (Expanded to all relax 3.2 versions.)
Jump to navigation Jump to search

Contents

Version 3 of relax

relax 3.3 series

relax 3.3.0

Descriptions

Download

The new relax versions can be downloaded from http://www.nmr-relax.com/download.html. If binary distributions are not yet available for your platform and you manage to compile the binary modules, please consider contributing these to the relax project (described in section 3.6 of the relax manual, http://www.nmr-relax.com/manual/relax_distribution_archives.html).


CHANGES file

Version 3.3.0
(3 September 2014, from /trunk)
https://sourceforge.net/p/nmr-relax/code/ci/3.3.0/tree/


Features
  • Huge speed ups for all of the relaxation dispersion models ranging from 1.452x to 163.004x times faster. The speed ups for the clustered spin analysis are far greater than for the single spin analysis.
  • Implementation of a zooming grid search algorithm for optimisation in all analyses. This includes the addition of the minimise.grid_zoom user function to set the zoom level. The grid width will be divided by 2zoom_level and centred at the current parameter values. If the new grid is outside of the bounds of the original grid, the entire grid will be translated so that it lies entirely within the original.
  • Increased the amount of user feedback for the minimise.grid_search user function. Now a comment for each parameter is included in the printed grid search setup table. This includes if the lower or upper bounds, or both, have been supplied and if a preset value has been used instead.
  • Expanded support for R 2D graph plotting in the relax_disp.plot_disp_curves user function as the X-axis can now be the ν1 value, the effective field ωeff, or the rotating frame title angle θ. And the plots are interpolation over the spin-lock offset.
  • Ability to optimise the R1 relaxation rate parameter in the off-resonance relaxation dispersion models.
  • Creation of the relax_disp.r1_fit user function for activating and deactivating R1 fitting in the dispersion analysis.
  • Better tab completion support in the prompt UI for Mac OS X users. For some Python versions, the Mac supplied libedit library is used rather than GNU readline. But this library uses a completely different language and hence tab completion was non-functional on these systems. The library difference is now detected and the correct language sent into libedit to activate tab completion.
  • Created the time user function. This is just a shortcut for printing out the output of the time.asctime() function.
  • The value.copy user function now accepts the force flag to allow destination values to be overwritten.
  • Expanded model nesting capabilities in the relaxation dispersion auto-analysis to speed up the protocol.
  • The spin-lock offset is now included in the spectra list GUI element for the relaxation dispersion analysis.
  • Creation of the relax_disp.r2eff_estimate user function for the fast estimation of R2eff/R values and errors when full exponential curves have been collected. This experimental feature uses linearisation to estimate the R2eff and I0 parameters and the covariance matrix to estimate parameter errors.
  • Gradients and Hessians are implemented for the exponential curve-fitting, hence all optimisation algorithms and constraint algorithms are now available for this analysis type. Using Newton optimisation instead of Nelder-Mead Simplex can save over an order of magnitude in computation time. This is also available in the relaxation dispersion analysis.
  • The minimisation statistics are now being reset for all analysis types. The minimise.calculate, minimise.grid_search, and minimise.execute user functions now all reset the minimisation statistics for either the model or the Monte Carlo simulations prior to performing any optimisation. This is required for both parallelised grid searches and repetitive optimisation schemes to allow the result to overwrite an old result in all situations, as sometimes the original chi-squared value is lower and the new result hence is rejected.
  • Large expansion of the periodic table information in the relax library to include all elements, the IUPAC 2011 standard atomic weights for all elements, mass numbers and atomic masses for all stable isotopes, and gyromagnetic ratios.
  • Significant improvements to the structure centre of mass calculations by using the new periodic table information - all elements are now supported and exact masses are now used.
  • Added a button to the spectra list GUI element for the spectrum.error_analysis user function. This is placed after the 'Add' and 'Delete' buttons and is used in the NOE, R1 and R2 curve-fitting and relaxation dispersion analyses.
  • RelaxErrors are now raised in the prompt or script UI if an old user function is called, printing out the names of the old and new user functions. This is for help in upgrading old scripts and is currently for the calc(), grid_search(), and minimise() user function calls.


Changes


Bugfixes


Links

For reference, the announcement for this release can also be found at following links:

Softpedia also has information about the newest relax releases:


relax 3.2 series

relax 3.2.3

Descriptions

Download

The new relax versions can be downloaded from http://www.nmr-relax.com/download.html. If binary distributions are not yet available for your platform and you manage to compile the binary modules, please consider contributing these to the relax project (described in section 3.6 of the relax manual, http://www.nmr-relax.com/manual/relax_distribution_archives.html).


CHANGES file

Version 3.2.3
(1 July 2014, from /trunk)
http://svn.gna.org/svn/relax/tags/3.2.3


Features
  • Improvements for a number of GUI elements used in the user function windows.
  • The ^[[?1034h escape code should now no longer be emitted by GNU readline on Linux systems.
  • Created the very basic structure.com user function for calculating the centre of mass. This is to simply allow an easy interface to the pipe_control.structure.mass.pipe_centre_of_mass() function.
  • Expansion of the REMARK section of the PDB file created for the internal structural object. This is visible when using the structure.write_pdb user function, as well as the many other user functions which create PDB files. The relax version as well as the file creation date are now recorded in the PDB file. This extra information should be very useful. Empty lines in the REMARK section improve the formatting.


Changes
  • Added proper sectioning to the release checklist document.
  • Added the upload script to the release checklist document.
  • Modified the Sequence GUI input element used for the user function list arguments. The first column is now of fixed with when titles are supplied. Previously when supplying titles, the width would be tiny and no text would be visible.
  • Added titles for all 3D coordinate user function arguments. This is for the Sequence GUI input element, and affects the frame_order.average_position, n_state_model.CoM and paramag.centre user functions.
  • The compilation of the C modules now respects the user defined environment. This is the patch from Justin attached to bug #22145. It has been modified to include a comment and remove a double empty line.
  • Bug fix for the compilation of the C modules now respects the user defined environment. The problem was that on Mac OS X (as well as other systems), that these environmental variables were not defined and hence the scons commands would all fail with a KeyError and traceback. Now the keys in the os.environ dictionary are being searched for before they are set.
  • Fix for the wxPython link in the installation chapter of the manual. This was pointing to the scipy website for some reason.
  • Changed the Python readline link for MS Windows in the installation chapter of the manual. This now points to https://pypi.python.org/pypi/pyreadline as the iPython link is broken.
  • Implemented system test Relax_disp.test_bug_22146_unpacking_r2a_r2b_cluster. This is to catch the wrong unpacking of R2A0 and R2B0 when performing a clustered full dispersion model analysis. Bug #22146 Unpacking of R2A0 and R2B0 is performed wrong for clustered "full" dispersion models.
  • Extended system test Relax_disp.test_bug_22146_unpacking_r2a_r2b_cluster for B14 full model. This is to catch the wrong unpacking of R2A0 and R2B0 when performing a clustered full dispersion model analysis. Bug #22146 Unpacking of R2A0 and R2B0 is performed wrong for clustered "full" dispersion models.
  • Extended system test Relax_disp.test_bug_22146_unpacking_r2a_r2b_cluster for NS CPMG 2-site 3D full model. This is to catch the wrong unpacking of R2A0 and R2B0 when performing a clustered full dispersion model analysis. Bug #22146 Unpacking of R2A0 and R2B0 is performed wrong for clustered "full" dispersion models.
  • Extended system test Relax_disp.test_bug_22146_unpacking_r2a_r2b_cluster for NS CPMG 2-site star full model. This is to catch the wrong unpacking of R2A0 and R2B0 when performing a clustered full dispersion model analysis. Bug #22146 Unpacking of R2A0 and R2B0 is performed wrong for clustered "full" dispersion models.
  • Added synthetic data generator script which created the data to test against. Bug #22146 Unpacking of R2A0 and R2B0 is performed wrong for clustered "full" dispersion models.
  • Split system test Relax_disp.test_bug_22146_unpacking_r2a_r2b_cluster up in different tests. A setup function which is: setup_bug_22146_unpacking_r2a_r2b_cluster(self, folder=None, model_analyse=None): And then the tests: test_bug_22146_unpacking_r2a_r2b_cluster_B14 test_bug_22146_unpacking_r2a_r2b_cluster_CR72 test_bug_22146_unpacking_r2a_r2b_cluster_NS_3D test_bug_22146_unpacking_r2a_r2b_cluster_NS_STAR. Bug #22146 Unpacking of R2A0 and R2B0 is performed wrong for clustered "full" dispersion models.
  • Modified profiling script to get closer to the implementation in relax. An additional test function is setup to figure out how to reshape the numpy arrays in the target function. Bug #22146 Unpacking of R2A0 and R2B0 is performed wrong for clustered "full" dispersion models.
  • Updated profiling text for CR72 model. Now it is tested for 3 fields. This is related to: Task #7807: Speed-up of dispersion models for Clustered analysis.
  • Added searching for environment variable PYTHON_INCLUDE_DIR if Python.h is not found in standard Python library. This can be very handsome, if one has a Python virtual environment for multiple users. This relates to the wiki page: http://wiki.nmr-relax.com/Epd_canopy.
  • The lib.compat.norm() replacement function for numpy.linalg.norm() now handles no axis argument. This is to allow the function to be used in all cases where numpy.linalg.norm() is used, while providing compatibility with the axis argument and all numpy versions.
  • Fix for the scons target for compiling the relax manual when using a repository checkout copy. The method for compiling the relax manual was calling the version.revision() function, however this has been replaced a while ago by the version.repo_revision variable.
  • Created two unit tests for the lib.io.file_root() function. The second of the tests demonstrate a failure of the function if multiple file extensions are present.
  • Lowered χ2 value test in system test Relax_disp.test_bug_22146_unpacking_r2a_r2b_cluster_NS_STAR. This is due to the data produced on 32 bit machine, and tested on 64 bit machines. The error was: AssertionError: 2.4659455670347743e-05 != 0.0 within 7 places. The reason for this is due to truncation artifacts.
  • Fix for wrong path testing of Python.h. Python.h would be in PYTHON_PREFIX/include/pythonX.Y/Python.h and not in PYTHON_PREFIX/include/Python.h.
  • Better handling of the control-C keyboard interrupt signal in the relax test suite. This includes two changes. The Python 2.7 and higher unittest.installHandler() function is now called, when present, to terminate all tests using the unittest module control-C handler. The second change is that the keyboard interrupt signal is caught in a try-except statement, a message printed out, and the tests terminated. This should be an improvement for all systems.
  • Adding last profiling information for model CR72.
  • Added system test for model LM63 3-site. According to results folder in test_suite/shared_data/dispersion/Hansen/relax_results/LM63 3-site. This should pass, but it doesn't.
  • Created an initial Relax_disp.test_lm63_3site_synthetic system test. This should have been set up a long time ago. It uses the synthetic noise-free data in the test_suite/shared_data/dispersion/lm63_3site directory which was created for a system test but never converted into one. The test still needs modifications to allow it to pass.
  • Modifications for the Relax_disp.test_lm63_3site_synthetic system test. The r2eff_values.bz2 saved state file has been updated, as it was too old to use in the test. The test has also had a typo bug fixed and the data pipe name updated. The test now also checks all of the optimised values.
  • Removed system test test_hansen_cpmg_data_to_lm63_3site. This was a temporary implementation and has been replaced with system test Relax_disp.test_lm63_3site_synthetic.
  • Fixes for all of the relaxation dispersion system tests which were failing with the new minfx code. Due to the tuning of the log barrier constraint algorithm in minfx in the commit at http://article.gmane.org/gmane.science.mathematics.minfx.scm/25, many system tests needed to be slightly adjusted. Two of the Relax_disp.test_tp02_data_to_* system tests were also failing as the optimisation can no longer move out of the minimum at pA = 0.5 for one spin (due to the low quality grid search in the auto-analysis).
  • Updated the release checklist document for the new 1.0.7 release of minfx.
  • Fixes for the Relax_disp.test_hansen_cpmg_data_missing_auto_analysis system test. The pA parameter is no longer tested for one spin as it moves to random values on different operating systems and 32 vs. 64-bit systems. This is because this spin experiences no exchange, both Δω and kex are zero.
  • Decreased the value checking precision in the Relax_disp.test_hansen_cpmg_data_to_lm63 system test. This is to allow the test to pass on certain operating systems and 32-bit systems.
  • Modified the precision of the output from the relax_disp.sherekhan_input user function. This is simply to allow the Relax_disp.test_sod1wt_t25_to_sherekhan_input system test to pass on certain 32-bit systems, as the float output to 15 decimal places is not always the same. This system test has been updated for the change.
  • Modified the Relax_disp.test_sprangers_data_to_mmq_cr72 system test to pass on certain systems. This test fails on 32-bit Linux (and probably other systems as well). To fix the test, the kex values are all divided by 100 before checking them to 4 decimal places of accuracy.
  • Improved how the relax installation path is determined in the status object. If the path cannot be found, the current working directory is then checked if it is where relax is installed. This is needed when importing modules outside of relax.
  • Hack to permanently eliminate the ^[[?1034h escape code being produced on Linux systems. This is produced by importing the readline module. The escape code will be sent to STDOUT every time relax is executed, so it will be present in all log files. The problem is the TERM environmental variable being set to 'xterm'. The hack simply sets TERM to an empty string.
  • More hacks for permanently eliminating the ^[[?1034h escape code being produced on Linux systems. This is a nasty feature of the GNU readline library. It is now also turned off in the dep_check module, suppressing ^[[?1034h in Python scripts which import only parts of relax.
  • Numpy version 1.6 or higher is now required to be able to run relax. This follows from the series of messages: http://www.mail-archive.com/relax-devel@gna.org/msg06288.html, http://www.mail-archive.com/relax-devel@gna.org/msg06289.html, http://www.mail-archive.com/relax-devel@gna.org/msg06327.html, and http://www.mail-archive.com/relax-devel@gna.org/msg06335.html. If too many users complain, maybe this change can be reverted later. This minimal numpy version is needed for many of the speed ups going in the relaxation dispersion and frame order analyses. It is required for the numpy ufunc out arguments and for the numpy.eigsum() function. These will likely be used in other analyses in the future for improving the speed of relax, so it might affect users of other analyses later on.
  • Updated the numpy minimal dependency in the installation chapter of the manual to version 1.6.
  • Added better epydoc sectioning to the lib.dispersion.ns_cpmg_2site_expanded module docstring. This is to better separate the original scripts used to document the code evolution.
  • Empty lines are now handled by the lib.structure.pdb_write.remark() function. By supplying the remark as None, empty lines can now be created in the REMARK section of a PDB file. This can be used for nicer formatting.
  • Fixes for the Diffusion_tensor system tests due to the recent PDB file changes. Prior to the comparison of the generated PDB files, all REMARK PDB lines are now stripped out.
  • Fixes for all system tests failing due to the expanded and improved PDB REMARK section. The system tests now remove all REMARK records prior to comparing file contents. The special strip_remarks() system test method has been created to simplify the stripping process.
  • Fix for the software verification tests. The recent expansion and improvements of the REMARK records created by the internal structural object PDB writing method imported the relax version to place this information into the PDB files. However this breaks the relax library design, as shown by the verification tests. Instead the relax version information is being taken from the lib.structure.internal.object.RELAX_VERSION variable. This defaults to None, however the version module now sets this variable directly when it is imported so that it is always set to the current relax version when running relax.
  • General Python 3 fixes via the 2to3 script.
  • Removed the lib.compat.sorted() function which was providing Python2.3 compatibility. For a while now, relax has been unable to run on Python versions less than 2.5. Therefore there is no use for having this replacement function for Python ≤ 2.3 which was being placed into the builtins module.
  • Python 3 fixes for the entire codebase using the 2to3 script. The command used was: 2to3 -j 4 -w -f xrange .
  • The internal structural object add_molecule() and has_molecule() methods are now model specific. This allows for finer control of structural object.
  • Created the new lib.structure.files module. This currently contains the single find_pdb_files() function which will be used to find all *.pdb, *.pdb.gz and *.pdb.bz2 versions of the PDB file in a given path.
  • Fix for the breakage of the relax help system. This was reported at http://thread.gmane.org/gmane.science.nmr.relax.devel/6481. The problem was that the TERM environmental variable was turned off to avoid the GNU readline library on Linux systems emitting the ^[[?1034h escape code. See the message at http://thread.gmane.org/gmane.science.nmr.relax.devel/6481/focus=6489 for more details. However the Python help system obviously requires this environmental variable. Now only if the TERM variable is set to 'xterm' will it be reset, and to 'linux' instead of the blank string "". This does not affect any relax releases.


Bugfixes


Links

For reference, the announcement for this release can also be found at following links:

Softpedia also has information about the newest relax releases:


relax 3.2.2

Descriptions

Download

The new relax versions can be downloaded from http://www.nmr-relax.com/download.html. If binary distributions are not yet available for your platform and you manage to compile the binary modules, please consider contributing these to the relax project (described in section 3.6 of the relax manual, http://www.nmr-relax.com/manual/relax_distribution_archives.html).


CHANGES file

Version 3.2.2
(5 June 2014, from /trunk)
http://svn.gna.org/svn/relax/tags/3.2.2


Features
  • Large speedups of all analytical relaxation dispersion models by converting the R2eff calculations and value error checking from single values to numpy arrays.
  • Edge cases where function failures occur are now properly handled for all analytical relaxation dispersion models.
  • Completion of the frame_order.pdb_model user function backend for the frame order PDB representation.
  • relax will now detect when files with *.gz or *.bz2 file extensions are being created and automatically gzip or bzip compress the file.


Changes
  • Small speed up for all the isotropic cone and pseudo-elliptic cone frame order models. The vector length calculation for the numeric PCS integration has been simplified and shifted outside of a loop to take advantage of the speed of numpy.
  • All three file arguments for the pymol.frame_order user function are now optional.
  • Updated all the API documentation links in the dispersion chapter of the manual. These were pointing to http://www.nmr-relax.com/api/3.1/ whereas they should now be point to http://www.nmr-relax.com/api/3.2/.
  • Modified a printout in the 'devel_scripts/code_validator' script. This is to clarify that the first method of a class does not need two preceding empty lines.
  • Shifted some functions from lib.structure.geometric into their own modules. The angles_regular() and angles_uniform() functions are now in the lib.structure.angles module, and the get_proton_names in lib.structure.conversion.
  • Deletion of the pipe_control.structure.main.create_cone_pdb() function. This is only used in the frame order analysis, but has been made redundant by the lib.structure.represent.cone.cone() function.
  • Completed the frame_order.pdb_model user function backend for the frame order PDB representation. Most of this backend, including the axes and cone representations, had been broken for quite a while and were being skipped with an early return statement. This has now been made functional and a few fixes have been made. For the 'rotor' and 'free rotor' model, the neg_cone argument is now ignored so that only one model is produced in the final frame order PDB representation file. For all other models, the rotor representation is no longer centred to the point on axis closest to the centre of mass, as the pivot is unambiguously defined. The rotor representation has also been made larger in these models so that it is outside of the cone, and the propeller blades are now staggered.
  • Modified py_type from "list" to "float_array" in uf_object type in user function dx.map. Bug #22035 The dx.map user function is broken in the GUI.
  • Added py_type "list_val_or_list_of_list_val" to be handled in GUI uf_objects. Bug #22035 The dx.map user function is broken in the GUI.
  • Modified the frame order constraints so that coneθx ≤ coneθy. The linear_constraints() function docstring has been updated to include this constraint.
  • Set dim=4 when setting chi surface level in user function dx.map.
  • Fix for the n_state_model.cone_pdb user function for the recent internal structural object changes. The cone arguments should now be called cone_obj.
  • Renamed the relax_disp.set_grid_r20_from_min_r2eff user function to relax_disp.r20_from_min_r2eff. This follows from the proposal at http://thread.gmane.org/gmane.science.nmr.relax.devel/5957.
  • Modification to the Sequence_2D GUI element used for some user function windows. The selection_win_show() method has been redefined, as the parent method from the Sequence element is specific for the 1D sequence module. The open_dialog() method has also been modified to use the new selection_win_show(), as well as the parent Sequence class selection_win_data() method.
  • Created the User_functions.test_structure_rotate GUI tests. This is to catch bug #22100, the rotation argument for the structure.rotate user function cannot be changed in the GUI, as an AttributeError is raised.
  • Moved py_type "list_val_or_list_of_list_val" to 2D sequence types.
  • Added dim dimensions to match the {x, y, z} positions for GUI input in user function dx.map.
  • Modified the User_functions.test_structure_rotate GUI test to change and check the rotation matrix.
  • Some more fixes for the User_functions.test_structure_rotate GUI test. The open_dialog() method cannot be used, as it deletes the window at the end. Instead the selection_win_show() and selection_win_data() method combination is used.
  • Expanded the User_functions.test_structure_rotate GUI test. This is to more extensively check the 'float_matrix' user function argument type in the GUI.
  • Modified the dim dimensions to (None, 3) to allow the user to change number of points in the GUI. This is for the user function dx.map.
  • Simplified the User_functions GUI tests. The exec_uf_pipe_create() method has been created to simplify the data pipe creation in the tests.
  • Expanded the User_functions.test_structure_rotate GUI test. The rotation matrix argument checks for the Sequence_2D GUI element have been expanded to check that setting nothing (blank element) returns nothing (None). The other checks have also been slightly modified.
  • Expanded the User_functions.test_structure_rotate GUI test to catch more problems. Now the rotation matrix value in the user function window is set to a series of invalid values to test if the Sequence_2D GUI element will handle the rubbish input. This is to mimic user errors.
  • Created the is_list() and is_list_of_lists() functions for the lib.check_types module.
  • Clean up of the User_functions.test_structure_rotate GUI test. The invalid value check is simpler and the Sequence_2D GUI object return value is now checked to be None.
  • Expanded the User_functions.test_structure_rotate GUI once more. This time the setting of invalid values in the Sequence_2D element itself is now checked. For example for the rotation matrix of the structure.rotate user function, if a matrix element is set to a string, a NameError is raised.
  • Created the User_functions.test_dx_map GUI test. This extensively checks the 'point' argument for the dx.map user function GUI window. This is to catch bug #22102, the point argument of the dx.map user function being incorrect in the GUI.
  • Modified the User_functions.test_dx_map GUI test to catch another problem with the Sequence_2D element.
  • Fixes for the frame order PDB presentation in the frame_order.pdb_model user function backend.
  • Expanded the User_functions.test_dx_map GUI test once again. The new test is to set 2 valid points in the wizard, open and close the Sequence_2D window (twice), and check that the points come back.
  • Increased the width of the first column of the Sequence_2D GUI element for variable lists. This is so the column title "Number" will fit.
  • Added list titles for the dx.map user function point argument. This is so that the Sequence_2D GUI element will have column titles of 'X coordinate', 'Y coordinate', and 'Z coordinate'.
  • The self.variable_length flag is now used throughout the Sequence GUI element.
  • The self.variable_length flag is used in one more spot in the Sequence_2D GUI element.
  • Created the User_functions.test_structure_add_atom GUI test. This is used to check the operation of the Sequence GUI element via the 'pos' argument of the structure.add_atom user function. This is a list fixed to 3 elements.
  • Titles are now handled and set in the Sequence GUI element. The titles will replace the numbering of 1 onwards in the first column of the GUI element.
  • Small fix for switched indices in the new User_functions.test_structure_add_atom GUI test.
  • Modified the 'pos' argument of the structure.add_atom user function. The argument is now a list of fixed length of 3, and it has the titles 'X coordinate', 'Y coordinate', and 'Z coordinate' which are shown in the GUI.
  • Created the User_functions.test_spectrum_read_intensities GUI test to catch bug #22105. The problem is that a single file name is split up into many files when the file selection button is clicked, one for each character of the file name.
  • Fix for the User_functions.test_spectrum_read_intensities GUI test. A valid value was being checked as invalid.
  • Shifted all wildcards used in GUI file selection dialogs into the new user_functions.wildcard module. These have now all been standardised, and expanded to include more capitalisation combinations and to include more *.* options.
  • Created a file selection wildcard for use in the GUI for selecting peak lists. This is used in the four user functions which read peak lists.
  • Changed all *.* GUI file selection wildcards to *.
  • Huge speedup for model CR72. Task #7793 Speedup of dispersion models. The system test Relax_disp.test_cpmg_synthetic_cr72_full_noise_cluster changes from 7 seconds to 4.5 seconds. This is won by not checking single values in the R2eff array for math domain errors, but calculating all steps, and in one single round check for finite values. If just one non-finite value is found, the whole array is returned with a large penalty of 1e100. This makes all calculations be the fastest numpy array way.
  • Fix for system test test_cpmg_synthetic_dx_map_points. Task #7793 Speedup of dispersion models.
  • Critical fixes for system test Relax_disp.test_hansen_cpmg_data_missing_auto_analysis. Task #7793 Speedup of dispersion models. It is suspected that when relax have touched boundary values which made math domain errors, the error catching have created local minima or interfered with the simplex search algorithm.
  • Speedup of model TSMFK01. Task #7793 Speedup of dispersion models. This is won by not checking single values in the R2eff array for math domain errors, but calculating all steps, and in one single round check for finite values. If just one non-finite value is found, the whole array is returned with a large penalty of 1e100. This makes all calculations be the fastest numpy array way.
  • Huge speedup of model B14. Task #7793 Speedup of dispersion models. Time test for system tests: test_baldwin_synthetic 2.626s -> 1.990s, test_baldwin_synthetic_full 18.326s -> 13.742s. This is won by not checking single values in the R2eff array for math domain errors, but calculating all steps, and in one single round check for finite values. If just one non-finite value is found, the whole array is returned with a large penalty of 1e100. This makes all calculations be the fastest numpy array way.
  • Speedup of model TP02. Task #7793 Speedup of dispersion models. The change for running system test is: test_curve_type_r1rho_fixed_time 0.057s -> 0.049s, test_tp02_data_to_ns_r1rho_2site 10.539s -> 10.456s, test_tp02_data_to_tp02 8.608s -> 5.727s. This is won by not checking single values in the R array for math domain errors, but calculating all steps, and in one single round check for finite values. If just one non-finite value is found, the whole array is returned with a large penalty of 1e100. This makes all calculations be the fastest numpy array way.
  • Huge speedup for model TAP03. Task #7793 Speedup of dispersion models. The change for running system test is: test_tp02_data_to_tap03 13.869s -> 7.263s. This is won by not checking single values in the R array for math domain errors, but calculating all steps, and in one single round check for finite values. If just one non-finite value is found, the whole array is returned with a large penalty of 1e100. This makes all calculations be the fastest numpy array way.
  • Speedup of model MP05. Task #7793 Speedup of dispersion models. The change in system test is: test_tp02_data_to_mp05 10.750s -> 6.644s.
  • Speedup of model MMQ CR72. Task #7793 Speedup of dispersion models. Change in system test: test_sprangers_data_to_mmq_CR72 9.892s -> 4.121s.
  • Speedup for model M61. Task #7793 Speedup of dispersion models. Change in speed is: test_m61_data_to_m61 6.692s -> 3.480s.
  • Speedup of model LM63. Task #7793 Speedup of dispersion models. Change in system test was: test_hansen_cpmg_data_auto_analysis 13.731s -> 9.971s, test_hansen_cpmg_data_auto_analysis_r2eff 13.370s -> 9.510s, test_hansen_cpmg_data_to_lm63 3.254s -> 2.080s.
  • Speedup of model IT99. Task #7793 Speedup of dispersion models. Change in speed is: test_hansen_cpmg_data_auto_analysis 9.74s -> 8.330s, test_hansen_cpmg_data_to_it99 4.928s -> 3.138s.
  • Speedup of model DPL94. Task #7793 Speedup of dispersion models. Change in speed is: test_dpl94_data_to_dpl94 19.412s -> 4.427s.
  • Math-domain catching for model B14. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These can be found via the --numpy-raise function to the system tests. To make the code look clean, the class object "back_calc" is no longer being updated per time point, but is updated in the relax_disp target function in one go.
  • Math-domain catching for model CR72. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These can be found via the --numpy-raise function to the system tests. To make the code look clean, the class object "back_calc" is no longer being updated per time point, but is updated in the relax_disp target function in one go.
  • Math-domain catching for model: NS CPMG 2-site expanded. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These can be found via the --numpy-raise function to the system tests. To make the code look clean, the class object "back_calc" is no longer being updated per time point, but is updated in the relax_disp target function in one go.
  • Math-domain catching for model CR72. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These can be found via the --numpy-raise function to the system tests. The skipping of test when num_points > 0, is a bad implementation. If such a case should show, it is best to catch the wrong input for the calculations. This is best done with a check before running the calculations.
  • Math-domain catching for model TSMFK01. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These can be found via the --numpy-raise function to the system tests. To make the code look clean, the class object "back_calc" is no longer being updated per time point, but is updated in the relax_disp target function in one go.
  • Math-domain catching for model TP02. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These can be found via the --numpy-raise function to the system tests. To make the code look clean, the class object "back_calc" is no longer being updated per time point, but is updated in the relax_disp target function in one go.
  • Math-domain catching for model TAP03. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These can be found via the --numpy-raise function to the system tests. To make the code look clean, the class object "back_calc" is no longer being updated per time point, but is updated in the relax_disp target function in one go.
  • Math-domain catching for model DPL94. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These can be found via the --numpy-raise function to the system tests. To make the code look clean, the class object "back_calc" is no longer being updated per time point, but is updated in the relax_disp target function in one go.
  • Math-domain catching for model TAP03. Another check for division with 0 inserted.
  • Math-domain catching for model MP05. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These can be found via the --numpy-raise function to the system tests. To make the code look clean, the class object "back_calc" is no longer being updated per time point, but is updated in the relax_disp target function in one go.
  • Math-domain catching for model IT99. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These can be found via the --numpy-raise function to the system tests. To make the code look clean, the class object "back_calc" is no longer being updated per time point, but is updated in the relax_disp target function in one go.
  • Removed class object "back_calc" being updated per time point for model LM63. Task #7793 Speedup of dispersion models. To make the code look clean, the class object "back_calc" is no longer being updated per time point, but is updated in the relax_disp target function in one go.
  • Math-domain catching for model M61. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These can be found via the --numpy-raise function to the system tests. To make the code look clean, the class object "back_calc" is no longer being updated per time point, but is updated in the relax_disp target function in one go.
  • Math-domain catching for model MMQ CR72. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These can be found via the --numpy-raise function to the system tests. To make the code look clean, the class object "back_calc" is no longer being updated per time point, but is updated in the relax_disp target function in one go.
  • Align math-domain catching for model CR72 with trunk implementation. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. The catching of errors have to be more careful.
  • Align math-domain catching for model DPL94 with trunk implementation. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. The catching of errors have to be more careful.
  • Align math-domain catching for model IT99 with trunk implementation. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. The catching of errors have to be more careful.
  • Align math-domain catching for model LM63 with trunk implementation. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. The catching of errors have to be more careful.
  • Align math-domain catching for model M61 with trunk implementation. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. The catching of errors have to be more careful.
  • Align math-domain catching for model MP05 with trunk implementation. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. The catching of errors have to be more careful.
  • Align math-domain catching for model TAP03 with trunk implementation. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. The catching of errors have to be more careful.
  • Align math-domain catching for model TP02 with trunk implementation. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. The catching of errors have to be more careful.
  • Align math-domain catching for model TSMFK01 with trunk implementation. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. The catching of errors have to be more careful.
  • Removing unnecessary math-domain catching for model IT99. Task #7793 Speedup of dispersion models. The denominator is always positive.
  • Align math-domain catching for model NS CPMG 2-site expanded with trunk implementation. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. The catching of errors have to be more careful.
  • Modified unit tests demonstrating edge case 'no Rex' failures of the model NS CPMG 2-site expanded. This is to align with the current return of data in the disp_speed branch. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0.
  • Added 7 unit tests demonstrating edge case 'no Rex' failures of the model DPL94. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0.
  • Unit test _lib/test_ns_cpmg_2site_expanded.py copied to _/test_lm63.py. They are both of CPMG type.
  • Added 7 unit tests demonstrating edge case 'no Rex' failures of the model LM63. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0.
  • Unit test _lib/_dispersion/test_ns_cpmg_2site_expanded.py copied to _lib/_dispersion/b14.py. They are both of CPMG type, and can be re-used.
  • Added 7 unit tests demonstrating edge case 'no Rex' failures of the model B14. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0.
  • Removed unnecessary math domain checking in model B14. They are slowing down the code. There is now protection for edge cases, and a last final check, before returning values. That should be sufficient.
  • Unit test _lib/_dispersion/test_b14.py copied to _lib/_dispersion/test_CR72.py. They are both of CPMG type, and can be re-used.
  • Copied unit test _lib/_dispersion/* to be reused for other models.
  • Added 8 unit tests demonstrating edge case 'no Rex' failures of the model CR72. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0; kex = 1e5.
  • Added the 8th unit tests demonstrating edge case 'no Rex' failures of the model B14. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0; kex = 1e5.
  • Added the 8th unit tests demonstrating edge case 'no Rex' failures of the model LM63. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0; kex = 1e20.
  • Small fix for 8 unit tests demonstrating edge case 'no Rex' failures of the model 'ns cpmg_2site_expanded'. The comparison of R2eff is now divided into a special case for kex having large values.
  • Deleted unit test case for lm63 3site.
  • Added 8 unit tests demonstrating edge case 'no Rex' failures of the model M61. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0; kex = 1e20.
  • Added the 8th unit tests demonstrating edge case 'no Rex' failures of the model DPL94. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange:
  • Added 8 unit tests demonstrating edge case 'no Rex' failures of the model M61b. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0; kex = 1e20.
  • Math-domain catching for model M61b. Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These can be found via the --numpy-raise function to the system tests. To make the code look clean, the class object "back_calc" is no longer being updated per time point, but is updated in the relax_disp target function in one go.
  • Modified script to be able to run system test Relax_disp.xxx_test_m61b_data_to_m61b.
  • Added 8 unit tests demonstrating edge case 'no Rex' failures of the model IT99. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0; kex = 1e19.
  • Added 9 unit tests demonstrating edge case 'no Rex' failures of the model MMQ CR72. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0; kex = 1e5; ΔωH = 0.0.
  • Added 8 unit tests demonstrating edge case 'no Rex' failures of the model MP05. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0; kex = 1e20.
  • Added 8 unit tests demonstrating edge case 'no Rex' failures of the model TAP03. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0; kex = 1e20.
  • Added 8 unit tests demonstrating edge case 'no Rex' failures of the model TP02. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0; kex = 1e20.
  • Added 7 unit tests demonstrating edge case 'no Rex' failures of the model TSMFK01. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0.
  • Copied unit test test_b14.py to test_ns_cpmg_2site_3d.py.
  • Added 8 unit tests demonstrating edge case 'no Rex' failures of the model NS CPMG 2-site 3D. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0; kex = 1e7.
  • Modified unit tests demonstrating edge cases 'no Rex' failures of the model TP02. The catching of errors for off-resonance R models was implemented wrong. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This was pointed out in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5938. This is to implement catching of math domain errors, before they occur. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0; kex = 1e5.
  • Critical fix for the math domain catching of model TP02. The catching of errors for off-resonance R models was implemented wrong. This was pointed out in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5938.
  • Modified unit tests demonstrating edge cases 'no Rex' failures of the model DPL94. This was pointed out in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5938.
  • Modified unit tests demonstrating edge cases 'no Rex' failures of the model MP05. The catching of errors for off-resonance R models was implemented wrong. This was pointed out in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5938. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur.
  • Critical fix for the math domain catching of model MP05. The catching of errors for off-resonance R models was implemented wrong. This was pointed out in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5938.
  • Modified unit tests demonstrating edge cases 'no Rex' failures of the model TAP03. The catching of errors for off-resonance R models was implemented wrong. This was pointed out in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5938. And post http://article.gmane.org/gmane.science.nmr.relax.devel/5944. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models.
  • Critical fix for the math domain catching of model TAP03. The catching of errors for off-resonance R models was implemented wrong. This was pointed out in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5938. And post http://article.gmane.org/gmane.science.nmr.relax.devel/5944.
  • Modified unit tests demonstrating edge cases 'no Rex' failures of the model MMQ CR72. This was pointed out in post http://article.gmane.org/gmane.science.nmr.relax.devel/5940. And in post http://article.gmane.org/gmane.science.nmr.relax.devel/5946. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models.
  • Small fix for the math domain catching of model MMQ CR72. This was pointed out in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5940. And in post http://article.gmane.org/gmane.science.nmr.relax.devel/5946.
  • Various spacing fixes in unit test files _lib/_dispersion. This is the preparation for merging back disp_speed branch into trunk. This follows post http://article.gmane.org/gmane.science.nmr.relax.devel/5948. Usin the code validator script './devel_scripts/code_validator'.
  • Modified that unit tests having different r20a and r20b values is checking if the correct one is returned. This is the preparation for merging back disp_speed branch into trunk. This follows post http://article.gmane.org/gmane.science.nmr.relax.devel/5948.
  • Modified unit test to have standard population of pA = 0.95, and a correctly calculation of Δω in ppm to rad/s. This is related to Task #7793 Speedup of dispersion models.
  • Small fix in parameter calculation in unit test _dispersion/test_ns_cpmg_2site_expanded.
  • Increased max kex to value 1e18 for unit test of lin/ns_cpmg_2site_expanded.py.
  • Increased max kex to value 1e20 for unit test of lib/ns_cpmg_2site_3d.py.
  • Fix for looking for negative values, when all values where converted to positive in matrix in ns_cpmg_2site_3d.py. This is to implement catching of math domain errors, before they occur. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models.
  • Removed nested looping of returning back_calc in lib/ns_cpmg_2site_3d.
  • Removed the 8th unit test for model NS CPMG 2-site 3D. This was the catching of errors when kex = 1e20. The model cannot handle this situations, and we need to let it fail.
  • Removed the 8th unit test for model NS CPMG 2-site expanded. This was the catching of errors when kex has high values. The model cannot handle this situations, and we need to let it fail.
  • Fix for differences in system tests which are different from trunk. These were found with the command: diff -bur disp_speed/test_suite/ relax_trunk/test_suite/ | grep -v "Binary files" > diff.txt.
  • Converting back to having back_calc as a function argument to model B14. This is to clean up the API. There can be bo no partial measures/implementations in the relax trunk. The problem is, that many numerical models can't be optimised further, since they evolve the spin-magnetisation in a matrix. That spin evolution can't be put into a larger numpy array. This is related to Task #7793 Speedup of dispersion models.
  • Converting back to having back_calc as a function argument to model CR72: This is to clean up the API. There can be bo no partial measures/implementations in the relax trunk. The problem is, that many numerical models can't be optimised further, since they evolve the spin-magnetisation in a matrix. That spin evolution can't be put into a larger numpy array. This is related to Task #7793 Speedup of dispersion models.
  • Converting back to having back_calc as a function argument to model DPL94: This is to clean up the API. There can be bo no partial measures/implementations in the relax trunk. The problem is, that many numerical models can't be optimised further, since they evolve the spin-magnetisation in a matrix. That spin evolution can't be put into a larger numpy array. This is related to Task #7793 Speedup of dispersion models.
  • Converting back to having back_calc as a function argument to model IT99: This is to clean up the API. There can be bo no partial measures/implementations in the relax trunk. The problem is, that many numerical models can't be optimised further, since they evolve the spin-magnetisation in a matrix. That spin evolution can't be put into a larger numpy array. This is related to Task #7793 Speedup of dispersion models.
  • Converting back to having back_calc as a function argument to model LM63: This is to clean up the API. There can be bo no partial measures/implementations in the relax trunk. The problem is, that many numerical models can't be optimised further, since they evolve the spin-magnetisation in a matrix. That spin evolution can't be put into a larger numpy array. This is related to Task #7793 Speedup of dispersion models.
  • Converting back to having back_calc as a function argument to model M61: This is to clean up the API. There can be bo no partial measures/implementations in the relax trunk. The problem is, that many numerical models can't be optimised further, since they evolve the spin-magnetisation in a matrix. That spin evolution can't be put into a larger numpy array. This is related to Task #7793 Speedup of dispersion models.
  • Converting back to having back_calc as a function argument to model M61b: This is to clean up the API. There can be bo no partial measures/implementations in the relax trunk. The problem is, that many numerical models can't be optimised further, since they evolve the spin-magnetisation in a matrix. That spin evolution can't be put into a larger numpy array. This is related to Task #7793 Speedup of dispersion models.
  • Converting back to having back_calc as a function argument to model MMQ CR72: This is to clean up the API. There can be bo no partial measures/implementations in the relax trunk. The problem is, that many numerical models can't be optimised further, since they evolve the spin-magnetisation in a matrix. That spin evolution can't be put into a larger numpy array. This is related to Task #7793 Speedup of dispersion models.
  • Converting back to having back_calc as a function argument to model MP05: This is to clean up the API. There can be bo no partial measures/implementations in the relax trunk. The problem is, that many numerical models can't be optimised further, since they evolve the spin-magnetisation in a matrix. That spin evolution can't be put into a larger numpy array. This is related to Task #7793 Speedup of dispersion models.
  • Converting back to having back_calc as a function argument to model NS CPMG 2-site expanded. This is to clean up the API. There can be bo no partial measures/implementations in the relax trunk. The problem is, that many numerical models can't be optimised further, since they evolve the spin-magnetisation in a matrix. That spin evolution can't be put into a larger numpy array. This is related to Task #7793 Speedup of dispersion models.
  • Converting back to having back_calc as a function argument to model TAP03. This is to clean up the API. There can be bo no partial measures/implementations in the relax trunk. The problem is, that many numerical models can't be optimised further, since they evolve the spin-magnetisation in a matrix. That spin evolution can't be put into a larger numpy array. This is related to Task #7793 Speedup of dispersion models.
  • Converting back to having back_calc as a function argument to model TP02. This is to clean up the API. There can be bo no partial measures/implementations in the relax trunk. The problem is, that many numerical models can't be optimised further, since they evolve the spin-magnetisation in a matrix. That spin evolution can't be put into a larger numpy array. This is related to Task #7793 Speedup of dispersion models.
  • Converting back to having back_calc as a function argument to model TSMFK01. This is to clean up the API. There can be bo no partial measures/implementations in the relax trunk. The problem is, that many numerical models can't be optimised further, since they evolve the spin-magnetisation in a matrix. That spin evolution can't be put into a larger numpy array. This is related to Task #7793 Speedup of dispersion models.
  • Created the lib.compat.norm() compatibility function for numpy.linalg.norm(). For numpy 1.8 and higher, the numpy.linalg.norm() function has introduced the 'axis' argument. This is an incredibly fast way of determining the norm of an array of vectors. This is used by the frame order analysis. However for older numpy versions, this causes the frame order analysis, and many corresponding system and GUI tests to fail. Therefore this new lib.compat.norm() function has been designed to default to numpy.linalg.norm() if the axis argument is supported, or to switch to the much slower numpy.apply_along_axis(numpy.linalg.norm, axis, x) call which is supported by older numpy.
  • The frame order analysis now uses the lib.compat.norm() replacement for numpy.linalg.norm(). This is to allow for the axis argument on numpy versions before version 1.8, though these older versions will result in slower optimisation of the frame order models.
  • The built in Python range() function is no longer being replaced by xrange(). Replacing builtin.range() with builtin.xrange() on Python 2 was causing problems with Python site-packages which were not Python 3 compliant. This includes old numpy versions. The original overwriting of range() with xrange() was for both speed and memory conservation. However profiling the system tests, the time for all tests did not change significantly. This change may cause problems in certain places in relax on memory constrained computer systems, so it may need to be reverted in the future.
  • The lib.io.open_write_file() function now automatically determines the compression type. This is used by many user functions which create files. The end result for a user is that if they supply a '.gz' or '.bz2' file extension, a gzipped or bzipped file will be produced.
  • Removal of the docstring text wrapping in the lib.io module.
  • Expanded and improved the docstring for the relax_disp.r20_from_min_r2eff user function. This follows from http://thread.gmane.org/gmane.science.nmr.relax.devel/5957. The documentation now covers a number of the uses for this user function. The text has also been lightly edited. To fit all the text into the GUI user function window, the size of the dialog and the text high settings have been changed.
  • Large improvements for the detection of cross-compilation on Mac OS X systems. The tests for different architecture support now follows the ideas discussed in the post http://thread.gmane.org/gmane.science.nmr.relax.devel/5785/focus=5820. In summary, for each architecture a simple C file is created, compiled with 'gcc -arch xyz', and the resultant binary file tested. To support 64-bit compilation on 32-bit systems, all previously successful architectures are also included in the gcc command. The change allows the 'ppc64' architecture to be reintroduced.
  • Fixed the docstring for the det_arch() method of the sconstruct script. This is for the true cross-compilation detection on Mac OS X.


Bugfixes
  • Fix for the lib.geometry.lines.closest_point_ax() function for when the two points are the same. If the point on the line and point in the 3D space are the same, then this function used to return an array of NaN values. This situation is now caught and the point in the 3D space is returned.
  • Fix for the heterogen section of the internal structural object write_pdb() method. A number of checks were performed to see if the PDB heterogen records were the same for all structures, but this is meaningless as the structures can of course be different.
  • Fixes for the lib.structure.represent.cone module. The function arguments named 'cone' have been renamed to 'cone_obj' so that they do not clash with the cone() function in the module namespace.
  • Fix for the lib.structure.geometric.generate_vector_residues() function. The atom numbers are no longer read from the internal structural object, as these are not reliable. If another geometric representation exists in the object, then the atom numbers could be None. Or loading structures from multiple PDB files can cause the numbering to be repeated or out of order.
  • Fix for the frame_order.pdb_model user function for the rotor models. The rotor axis is no longer defined by spherical angles and therefore needs to be recreated using the create_rotor_axis_alpha() rather than create_rotor_axis_spherical() function from lib.frame_order.rotor_axis.
  • Partial fix for bug #22100, the rotation argument for the structure.rotate user function cannot be changed in the GUI, as an AttributeError is raised. The append_row() method call has been replaced by the correct add_element() call.
  • Bug fix for the Sequence_2D GUI element. This is used for the user function windows in the GUI for setting lists of lists or matrices. The GUI element GetValue() method will now return None if nothing is set. This prevents a list of lists of None being added to the main user function window.
  • Fixes for the Sequence and Sequence_2D GUI elements for handling invalid input data. These elements used by the user function windows previously raised all sorts of errors if the data was not what they expected (lists or lists of lists respectively). These situations are now caught and the input data is ignored, so blank Sequence and Sequence_2D elements are presented to the user.
  • Bug fix for the Sequence_2D GUI element. This is used for handling list of lists user function arguments in the user function GUI windows. The setting of invalid values directly in the Sequence_2D GUI element is now detected. These values are now replaced with None.
  • Fix for bug #22102, the point argument of the dx.map user function failing in the GUI. The Sequence_2D GUI element used for all list of lists arguments in the user function GUI windows now correctly handles variable length lists. The first column which shows a count of the elements is now properly taken into account in the SetValue(), GetValue() and add_item() methods, via a new self.offset variable. The self.variable_length variable has also been fixed so it is not overwritten by the parent Sequence GUI element.
  • Bug fixes for the Sequence GUI element used for lists in the user function windows. Invalid values input into the Sequence GUI window are now ignored rather than raising different types of error. And invalid input lists for fixed dimension arguments are also ignored. This allows the User_functions.test_structure_add_atom GUI test to pass.
  • Bug fix for the lib.arg_check.is_float_object() function. The dim argument can sometimes be an integer rather than a tuple, but this was not handled by the function. Now integer dim arguments are pre-converted to lists before performing all the checks.
  • Fix for bug #22105, the failure spectrum.read_intensities GUI user function whereby a file name is turned into lists of characters. A few changes were made to allow the Selector_file_multiple GUI element to operate correctly.
  • Critical fix for the math domain catching of model LM63. This was discovered with the added 7 unit tests demonstrating edge case 'no Rex' failures. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur.
  • Critical fix for the math domain catching of model B14. This was discovered with the added 7 unit tests demonstrating edge case 'no Rex' failures. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur.
  • Critical fix for the math domain catching of model CR72. This was discovered with the added 8 unit tests demonstrating edge case 'no Rex' failures. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. When kex is large, e.g. kex = 1e5, the values of etapos = eta_scale * sqrt(Psi + sqrt_psi2_zeta2) / cpmg_frqs will exceed possible numerical representation. The catching of these occurrences needed to be re-written.
  • Critical fix for the math domain catching of model B14. This was discovered with the added 8 unit tests demonstrating edge case 'no Rex' failures. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. When kex is large, e.g. kex = 1e5, "nan" values where produced, which were replaced with 1e100. The catching of these occurrences needed to be re-written.
  • Critical fix for the math domain catching of model CR72. Removed the test for kex ≥ 1e5. This catching should rather be performed on the math functions instead.
  • Critical fix for the math domain catching of model B14. Removed the test for kex ≥ 1e5. This catching should rather be performed on the math functions instead. In this case, it is the catching of sinh(), not evaluating values above 710.
  • Critical fix for the math domain catching of model MMQ CR72. This was discovered with the added 9 unit tests demonstrating edge case 'no Rex' failures. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur.
  • Critical fix for the math domain catching of model IT99. This was discovered with the added 8 unit tests demonstrating edge case 'no Rex' failures. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. The catching of these occurrences needed to be re-written.
  • Critical fix for the math domain catching of model TAP03. This was discovered with the added 8 unit tests demonstrating edge case 'no Rex' failures. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur.
  • Critical fix for the math domain catching of model TP02. This was discovered with the added 8 unit tests demonstrating edge case 'no Rex' failures. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur.
  • Critical fix for the math domain catching of model TSMFK01. This was discovered with the added 8 unit tests demonstrating edge case 'no Rex' failures. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur. The catching of these occurrences needed to be re-written.
  • Critical fix for the math domain catching of model NS CPMG 2-site 3D. This was discovered with the added 8 unit tests demonstrating edge case 'no Rex' failures. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. This is related to Task #7793 Speedup of dispersion models. This is to implement catching of math domain errors, before they occur.
  • Fix for bug #22112, the GUI failure when setting list values via the sequence windows, launched from user function windows fails on Mac OS X. The problem was two fold. First the Sequence and Sequence_2D windows from wx.Dialog should not be terminated via the Destroy() method, as wx.Dialog.Destroy() appears to be horribly broken on Macs.
  • Another fix for bug #22112, the GUI failure when setting list values via the sequence windows, launched from user function windows fails on Mac OS X. This change is for the multiple file selection window and matches the previous change by replacing the Mac OS X fatal wx.Dialog.Destroy() call with wx.Dialog.Close().
  • Fix for the relax start up detection of missing Python packages. The dep_check module is now imported first, as it used to be. This is required to check if all required modules are installed and to present understandable messages to the user rather than cryptic ImportError messages with tracebacks.
  • Fix for bug #22033, the inability to use other optimisation algorithms in the dispersion analysis. As mentioned in comment #2, the solution is to raise a RelaxError explaining that only 'simplex' optimisation is possible for the dispersion analysis as the gradients are not derived and implemented in relax.


Links

For reference, the announcement for this release can also be found at following links:

Softpedia also has information about the newest relax releases:


relax 3.2.1

Descriptions

Download

The new relax versions can be downloaded from http://www.nmr-relax.com/download.html. If binary distributions are not yet available for your platform and you manage to compile the binary modules, please consider contributing these to the relax project (described in section 3.6 of the relax manual, http://www.nmr-relax.com/manual/relax_distribution_archives.html).


CHANGES file

Version 3.2.1
(23 May 2014, from /trunk)
http://svn.gna.org/svn/relax/tags/3.2.1


Features
  • N/A


Changes
  • Punctuation fixes throughout the CHANGES document.
  • Modified system test Relax_disp.test_cpmg_synthetic_ns3d_to_cr72 to catch bug #22017: LinAlgError, for all numerical CPMG models. System test was renamed from test_cpmg_synthetic_cr72 to test_cpmg_synthetic_ns3d_to_cr72, to reflect which model create the data and which model fits the data.
  • Modified cpmg_synthetic script to first create all time structures before doing back-calculation. Bug #22017: LinAlgError, for all numerical CPMG models. The numerical models need all time points which are defined in setup to be present when calculating.
  • Renamed system test to test_cpmg_synthetic_ns3d_to_cr72_noise_cluster. The model that creates the data has been changed to numerical model. Bug #22017: LinAlgError, for all numerical CPMG models.
  • Implemented system test Relax_disp.test_cpmg_synthetic_ns3d_to_b14. Bug #22021: model B14 shows bad fitting to data. This is to catch model B14 showing bad fitting behaviour.
  • Parameter precision increase for system test Relax_disp.test_baldwin_synthetic. The correct implementation of the trigonometric functions allow for higher precision. Bug #22021: model B14 shows bad fitting to data. Duplicate line codes were also removed.
  • Code cleanup in system test Relax_disp.test_baldwin_synthetic_full. Bug #22021: model B14 shows bad fitting to data. The precision could also be increased by 1 digit.
  • Code cleanup in system test Relax_disp.test_baldwin_synthetic. Bug #22021: model B14 shows bad fitting to data. Removing many unnecessary lines of code.
  • Added 7 unit tests demonstrating edge case 'no Rex' failures of the NS CPMG 2-site expanded model. This follows from the ideas in the post http://article.gmane.org/gmane.science.nmr.relax.devel/5858. These tests cover all parameter value combinations which result in no exchange: Δω = 0.0; pA = 1.0; kex = 0.0; Δω = 0.0 and pA = 1.0; Δω = 0.0 and kex = 0.0; pA = 1.0 and kex = 0.0; Δω = 0.0, pA = 1.0, and kex = 0.0. Such tests should be replicated for all dispersion models.
  • Created the Structure.test_bug_22069_structure_delete_helix_attribute system test. This is to catch bug #22069, the failure of the structure.delete user function with "AttributeError: Internal instance has no attribute 'helices'".
  • Created the Structure.test_bug_22070_structure_superimpose_after_deletion system test. This is to catch bug #22070, the failure of the structure.superimpose user function after deleting atoms with structure.delete.
  • Added some checks to the Structure.test_bug_22070_structure_superimpose_after_deletion system test. These tests reveal the real problem - that the atoms of the second model have not been removed by the structure.delete user function.
  • Added git-svn support for the relax version information module. This allows the subversion revision number and repository URL to be displayed on program startup, so that it is stored in log files. This is very useful for debugging purposes.
  • Improvements for the git-svn support in the relax version module. Python 3 is now correctly handled and the URL is properly extracted from the git repository.
  • Improvement for the unit test printouts when run with the --time command line option. The full unit test name is now printed out, reverting to the old behaviour. However the shortened test names are preserved for the other test suite categories.
  • Created the test_ns_cpmg_2site_expanded_no_rex8() relaxation dispersion unit test. This is a demonstration, showing the NS CPMG 2-site expanded model with no exchange when kex = 1e5. I.e. when the motion is too fast for exchange to be observed. This test should be used for all dispersion models to make sure that they model this edge case correctly as well. This follows from http://article.gmane.org/gmane.science.nmr.relax.devel/5906.
  • Attempt at fixing bug #22071, the relax unit test and system test not functioning. The fix here is that the git commands to show the current subversion revision number only works when run from the relax base directory, or one of the subdirectories. This should now be fixed, as the pipe running the command will first 'cd' to the relax base directory.
  • Another attempt at fixing bug #22071, the relax unit test and system test not functioning. This time the complicated shell command "cd %s; git svn find-rev $(git rev-parse HEAD)" has been replaced with "cd %s; git svn info".
  • Changed most default dispersion parameter values to avoid edge cases where there is no exchange. The Δω parameters were all 0.0 and kex 1e5, both of which result in no exchange. If this is ever used for an optimisation starting point - which it never should, apart from development, test suite, and debugging purposes - then the optimisation algorithm will have a very hard time recovering. The pA parameter has been changed to 0.90 to set it to a reasonable value while still staying far away from the no exchange condition of pA = 1.0. This follows from http://article.gmane.org/gmane.science.nmr.relax.devel/5917.
  • Fixes for 3 dispersion system tests for the change in default parameter values. The default values are used in the auto-analysis in the test suite to avoid the grid search. The changed values affected the optimisation of two spins from Flemming Hansen's data located at test_suite/shared_data/dispersion/Hansen/, residue 4 used as an example of no exchange and residue 70 used as an example where data is only available at one field. The system test Relax_disp.test_set_grid_r20_from_min_r2eff_cpmg was also modified as it was directly checking these default values.
  • Fix for the Relax_disp.test_cpmg_synthetic_dx_map_points system test. This uses the default parameter values to start the optimisation, therefore the recent change away from edge case 'no Rex' values allows the parameter values stored in ds.dx_clust_val to be correctly optimised.
  • Speed up for the version module when using a repository copy of the code. The repository revision and URL and now stored as module variables, so that the 'svn info' and 'git svn info' commands are only run twice, once for the revision() function and once for the url() function.
  • Large speed up for the relax start up times for svn and git-svn copies of the relax repository. The 'svn info' and 'git svn info' commands are now only executed once when the version module is first imported. The revision() and url() functions have been merged into the repo_info() function and this is called when the module is imported. This repo_info() function stores the repository revision and URL as the version.repo_revision and version.repo_url module variables. It also catches if these variables are already set, so that multiple imports of the module do not cause the repository information to be looked up each time. Previously the revision() and url() functions where called every time a relax state or result file was created, hence for repository copies the 'svn info' or 'git svn info' commands were being called each time. The functions were also called for each interpreter object instantiated, and for each import of the version module.


Bugfixes


Links

For reference, the announcement for this release can also be found at following links:

Softpedia also has information about the newest relax releases:


relax 3.2.0

Descriptions

Download

The new relax versions can be downloaded from http://www.nmr-relax.com/download.html. If binary distributions are not yet available for your platform and you manage to compile the binary modules, please consider contributing these to the relax project (described in section 3.6 of the relax manual, http://www.nmr-relax.com/manual/relax_distribution_archives.html).


CHANGES file

Version 3.2.0
(20 May 2014, from /trunk)
http://svn.gna.org/svn/relax/tags/3.2.0


Features
  • Addition of the vector_angle() relax library function for calculating the signed or directional angle between two vectors.
  • Huge speed up of the interatom.define user function.
  • For improved feedback, a busy cursor is shown in the GUI when executing user functions.
  • The steady-state NOE auto-analysis now produces a 2D Grace plot of the reference and saturated spectra peak intensity values.
  • Complete redesign of the specific analyses backend, simplifying and cleaning up this internal API and making it easier for users to add completely new analysis types to relax.
  • Parametric reduction of the rotor frame order model, eliminating one redundant parameter hence simplifying optimisation.
  • Large improvement for the lib.software.grace module. The '*_err' and '*_bc' parameter names for the parameter error and back-calculated parameters respectively are now supported, allowing these values to be easily plotted.
  • Expansion of the value.set user function to handle parameters which consist of lists of values. The index argument has been added to allow the index of the list to be specified, and this is then propagated into the specific analysis API.
  • Improvements for the parameter definitions in all analysis types. This allows for better output in 2D plots and text files.
  • Implemented linear constraints for the frame order analysis. This uses the log-barrier constraint algorithm in the minfx library to provide constraints without requiring gradients.
  • Improved and expanded the relax command line options for debugging.
  • Full independence of the relax library so that it can be used outside of relax.
  • The addition of a relaxation dispersion user function for setting the R20 values to the minimum R2eff value.
  • Expanded capabilities for the relax_disp.sherekhan_input user function.
  • Implementation of the B14 and B14 full relaxation dispersion CPMG models for 2-site exchange for all time scales (from the new paper [Baldwin 2014] at http://dx.doi.org/10.1016/j.jmr.2014.02.023).
  • Large improvements to the relax HTML manual including fixes for URLs, bibliography entries, links, and tables.
  • Support for multiple point creation for the OpenDX chi-squared space mapping user function.
  • Automatic determination of reasonable initial contour levels for the OpenDX mapping user function.
  • Addition of a new chapter to the manual for the N-state model or ensemble analysis.
  • Creation of the new pymol.frame_order user function for visualising results.
  • Expansion of the Grace 2D data plotting capabilities.


Changes


Bugfixes


Links

For reference, the announcement for this release can also be found at following links:

Softpedia also has information about the newest relax releases:


relax 3.1 series

relax 3.1.7

relax 3.1.6

relax 3.1.5

relax 3.1.4

relax 3.1.3

relax 3.1.2

relax 3.1.1

relax 3.1.0

relax 3.0 series

relax 3.0.2

relax 3.0.1

relax 3.0.0

Version 2 of relax

relax 2.2 series

relax 2.2.5

relax 2.2.4

relax 2.2.3

relax 2.2.2

relax 2.2.1

relax 2.2.0

relax 2.1 series

relax 2.1.2

relax 2.1.1

relax 2.1.0

relax 2.0 series

relax 2.0.0

Version 1 of relax

relax 1.3 series

relax 1.3.16

relax 1.3.15

relax 1.3.14

relax 1.3.13

relax 1.3.12

relax 1.3.11

relax 1.3.10

relax 1.3.9

relax 1.3.8

relax 1.3.7

relax 1.3.6

relax 1.3.5

relax 1.3.4

relax 1.3.3

relax 1.3.2

relax 1.3.1

relax 1.3.0

relax 1.2 series

relax 1.2.15

relax 1.2.14

relax 1.2.13

relax 1.2.12

relax 1.2.11

relax 1.2.10

relax 1.2.9

relax 1.2.8

relax 1.2.7

relax 1.2.6

relax 1.2.5

relax 1.2.4

relax 1.2.3

relax 1.2.2

relax 1.2.1

relax 1.2.0

relax 1.0 series

relax 1.0.10

relax 1.0.9

relax 1.0.8

relax 1.0.7

relax 1.0.6

relax 1.0.5

relax 1.0.4

relax 1.0.3

relax 1.0.2

relax 1.0.1

relax 1.0.0

See also