0.9.9
- Remove admin-requirement during install (Windows)
- Support newer correlator.com .sin file format (experimental, #135)
- Add smart progress dialog for fitting (#155)
- Statistics: check "filename/title" by default (#151)
- Documentation: fix bad LaTeX commands (#163)
0.9.8
- Bugfixes:
  - Indexing error when saving sessions (#154)
  - Page number truncated in csv export (#159)
  - Export of csv files used incorrect normalization (#153)
  - Normalization parameter was not displayed in the 'Info' tool
0.9.7
- Second triplet time is now larger than first triplet time
  by default
- Remove a hack that causes a run through all pages
  e.g. when an average is created
- Bugfixes:
  - Opening sessions with user-defined models
  - Saving sessions with comments containing non-ASCII characters
  - Windows build: Graphical plot export was misconfigured
    -> added matplotlibrc patch in .spec file
0.9.6
- Bugfixes:
  - Fixed minor wx sizer problems for the tools
  - Fixed `AttributeError` in page.py if no weights are present
- New confocal fitting models (#111):
  - 3D+3D, 2D+2D, 3D+2D; no triplet
  - T+T+2D+2D, T+T+3D+2D; double triplet
  - T+3D+3D+3D, T+3D+3D+2D (#40, #59)
- Under the hood:
  - Separation of core and GUI modules
  - Include tests in distributions for PyPI
  - Improve automated testing on Windows and Mac OS
  - More constraint options for fitting
0.9.5
- Bugfixes
  - Closing the batch control window causes segfault bug (#142)
  - Closing page causes error when batch control is active (#143)
  - Plot normalization causes "Save Session" to fail (#144)
  - Plot normalization not loaded from session (#145)
0.9.4
- Batch control allows to select individual parameters (#108)
- Allow to exclude pages from batch fitting (#107)
- Bugfixes:
  - Fix `ValueError` in parameter display
  - Possibly fixed error with `yaml.safe_dump`
    on Mac OSx 10.8.5
  - Make sure background is lower than signal (#137)
0.9.3
- Fitting: migrate to lmfit
  - This introduces a new dependency for building PyCorrFit.
    (e.g. in Debian, the package "python-lmfit" is required)
  - Improved fitting behavior at parameter boundaries
  - Removed "Polak-Ribiere" fitting algorithm
  - Added "Sequential Linear Squares Programming" algorithm
- Heuristic fit (#109):
  - Detect parameters that are stuck during fitting
  - Fit each curve five times or less and check
    whether the fit converges.
  - If two diffusion time parameter exist in a model, always
    make sure that one parameter is the larger one. This
    feature can currently not be disabled (#110).
- Allow infinity ("inf" and "-inf") parameters for models
  and boundaries.
- New model: confocal T+T+3D+3D
- Bugfixes:
  - Sessions saved with 64bit Windows were not opened (#136)
  - Old sessions and "KeyError: 'chi2'"
  - Old session file extension was not recognized (#106)
0.9.2
- Bugfixes:
  - "Slider Simulation"/"Parm Range" broken (#133)
  - Computation of average intensity did not work
    correctly for unequally spaced traces
- Update .pt3 reader to version 8399ff7401
- Import traces of .pt3 files (experimental, #118)
  Warning: Absolute values for intensity might be wrong
0.9.1
- Tool 'Overlay curves': improve UI (#117)
- Tool 'Statistics view': improve UI (#113)
- Tool 'Trace view': display countrate (#121)
- Bugfixes:
  - Unicode errors in statistics tool (#131)
  - Load session errors with empty pages
0.9.0
- Improve parameter display (#52, #114)
- Display Chi2 on each page (#115)
- The displayed Chi2-value for non-weighted fits is now
  normalized to the expected values of the fit. The
  documentation has been updated accordingly.
- Add "All files" option in save dialogs (#97)
- Improved plot export dialog (#99)
0.8.9
- Improved support for "ALV-7004" files (#104)
- Increase resolution for image export
- Load weights from PyCorrFit csv files 
- Tool 'Overlay Curves': show cropped correlation curves
- Tool 'Trace view': increase size of window (#93)
- Tool 'Global fitting': remove forced, joint weights
- Session comment dialog: more intuitive behavior (#116)
- Improve plot export (#95)
- Bugfixes:
   - Weighted fits at borders of fit interval were
     computed incorrectly due to integer division
   - Fitting algorithms did not work (#94)
   - Creating averages did not work (#123)
   - ASCII errors in statistics tool (#112)
- Under the hood:
   - Introduce new classes: Correlation, Fit, Trace
   - Code cleanup and rewrite to support planned features
   - In some cases support older versions of NumPy
0.8.8
- Improved support for "ALV-7004" files
- If you install the GUI with pip, you now need to include
  the `GUI` requirement: `pip install pycorrfit[GUI]`.
  The GUI depends on matplotlib and wxPython which is not
  required for scripting with the pycorrfit module.
- Bugfix: missing version string on SuSe linux (#101)
- Under the hood:
   - Python entry point script replaces "bin/" script
   - Windows build system hosted by appveyor.com
   - MacOS X build system hosted by travis-ci.org
   - New builds use wxPython3 (#85)
   - Unicode support without `reload(sys)`
   - Error messages are more verbose
0.8.7
- Removed unused fitting parameter d_eva from model 6022 and
  secured backwards compatibility.
- Improved support for ALV700X (#92)
- Bugfix: Corrected false display of Unicode characters on Windows
- Under the hood:
   - Code cleanup with pyflakes
   - Repo cleanup (#98)
0.8.6
- Bugfix: Opening .fcs files with only one AC curve works now
- Zip files with measurements may now contain subfolders
- Improved pt3-file support from
  https://github.com/dwaithe/FCS_point_correlator (#89)
0.8.5
- Fixed bug that made it impossible to load data (#88)
- Exceptions are now handled by wxPython
- Under the hood:
   - Pythonic repository structure
   - Relative imports
   - Windows build machine is now Windows 7
   - Removed strict dependency on matplotlib
0.8.4
- Support for PicoQuant data file format
  Many thanks to Dominic Waithe (@dwaithe)
- Improved compatibility with Zeiss .fcs file format
- PyCorrFit is now dependent on Cython
- The module 'openfile' is now available from within Python
- Installer for Windows
0.8.3
- New .pcfs (PyCorrFit Session) file format (#60)
- Additional fitting algorithms: Nelder-Mead, BFGS, Powell, Polak-Ribiere (#71)
- Improvements
   - Massive speed-up when working with large data sets (#77)
   - Plot export: legend position and displayed parameters (#54)
   - Average tool: traces may now start at time points != 0
   - Statistics tool: display on smaller screens
   - ALV data files: updated parser to identify curve types and segment traces
   - Zeiss ConfoCor3 data files: some files could not be opened due to dummy AC curves
   - Models: default parameters were changed to prevent unstable fits
   - Software: notification dialogs for missing modules or other software
- Bugfixes
   - User could accidently clear a session (#65)
   - wxPython plotting problem on MacOSx (#64)
   - Statistics view: some parameters were duplicated (#76)
   - Caught zero-division warnings (models with triplet component)
   - Corrected x-axis scaling of statistics view and trace view
0.8.2
- The documentation has been thoroughly reworked
- The user is now warned if he does not have a TeX distribution installed
- Improvements:
   - Complete support for installing PyCorrFit with virtualenv and pip
     (This is documented in the wiki)
   - Statistics tool now displays average and SD (#43)
- Bugfix: TeX did not work on Ubuntu due to missing imports
0.8.1
- Thanks to Alex Mestiashvili for providing initial setup.py files
  and for debianizing PyCorrFit (@mestia)
- Thanks to Thomas Weidemann for his contributions to the documentation (@weidemann)
- Bugfixes
   - Some ConfoCor files were not imported
   - The cpp was not calculated correctly in case of background correction (#45)
   - Enabled averaging of single pages (#58)
   - Background correction for cross-correlation data is now computed (#46)
- Improvements of the user interface
   - The menus have been reordered (#47, #50)
   - The fitting panel has been optimized (#49)
   - the slider simulation got a reset button (#51)
   - The Help menu contains documentation and wiki (#56)
   - Model functions that are somehow redundant have been removed from the menu,
     but are still supported
   - The model doc strings were fully converted to Unicode
   - Several text messages were modified for better coherence
   - The background correction tool is more intuitive
   - Statistics panel improvements (#43)
   - Run information is included in the Data set title
   - The page counter starts at "1" instead of "0" (#44)
   - New handling of background correction (#46, #53)
0.8.0
- Filename/title of each tab now shows up in the notebook (#39)
- Statistics tool can plot parameters and page selection with the Overlay
  tool is possible (#31)
0.7.9
- Support for Mac OSx
- Enhancements:
   - Export file format (.csv) layout improved
   - Model function info text in UTF-8
   - Improved waring message when opening sessions from future versions
   - New feature lets user set the range for the fitting parameters
- Bugfixes:
   - Cleaned minor tracebacks and exceptions created by the frontend
   - Mac version now works as expected, but .app bundling failed
   - Latex plotting features now support more characters such as "[]{}^"
0.7.8
- Enhancements:
   - Averages can now be calculated from user-selected pages
   - Pages selected in the Overlay tool are now automatically set
     for computation of average and for global fitting
   - Source pages are now displayed in average title
   - Graph normalization with particle numbers is now possible
- Bugfixes:
   - Errors during fitting with weights equal to zero
   - Overlay tool displayed last curve after all pages have been removed
   - Global fit did not work with weights
- Session saving now uses 20 digits accuracy
- CSV export is now using tab-delimited data for easier Excel-import
- Added version checking for session management
0.7.7
- Fixed: Tools windows could not be closed (or moved on MS Windows)
- Fixed: .csv export failed in some cases where no weights were used
- Enhancement: The user is now asked before a page is closed
- Enhancement: Tool "Page Info" and in exported .csv files, variables and
               values are now separated by a tab stop instead of a "="
- Fixed: Opening a session with an empty page failed in some cases
- Fixed: Tool "Statistics" missed to output the column "filename/title"
         if that key value is empty - replaced empty strings with "NoName"
- Enhancement: Tool "Overlay" now asks the user to check kept curves
               instead of showing the curves to be removed
- Enhancement: Tool "Overlay" now has a "Cancel" button
0.7.6
- Improved handling
   - Tools are now sorted according to a standard work-flow
   - Renamed "Curve selection" to "Overlay tool" - this is more intuitive
   - Tools will now stay open or may be opened when there are no open pages (#25)
   - Filenames and runs are now displayed on each page (also added filename/title tag) (#23)
   - Notebook: moved close button to each tab to prevent accidental closing of tabs
- Improved tool "Statistics" (#21)
   - Fixed the case where "useless" data was produced - instead we write "NaN" data,
     removed warning message accordingly
   - Row-wise ordering according to page numbers (#22)
   - Column-wise ordering is now more intuitive
     (Fitted parameters with errors first)
   - Some columns are now checked by default
   - PyCorrFit remembers checked parameters for a page (not saved in session)
- Improved tool "Overlay" (#23)
   - New feature: Overlay shows run number of each file (upon import),
     the run (or index) of an experimental file is unique to PyCorrFit
   - Upon import, filenames and runs are displayed
   - In a session, the filename/title is displayed
- Web address of PyCorrFit changed
  from "fcstools.dyndns.org/pycorrfit" to "pycorrfit.craban.de"
- Minor bugfixes: Batch control, Global fitting, import dialog
0.7.5
 - Added model functions to documentation.
 - Weights from fitting are now exported in .csv files.
 - Rework of info panel for fitting
 - Cleared unintuitive behavior of session saving:
   The fitting parameters were read from the frontend. This could have
   led to saving false fit meta data.
 - During fitting, units are now displayed as "human readable" (#17).
 - Slider simulation now also uses human readable units (#17).
 - Secured support for Ubuntu 12.10 and 13.04
 - Fixed: new line (\n) characters for LaTeX plotting on Windows
0.7.4
 - New tool: Colorful curve selection
 - Import data: Curve selection possible
 - Average: Crop average according to current page.
 - Fixed: Page now displays Chi-squared of global fit.
 - Fixed: Chi-squared and parameters of global fitting are now stored in sessions.
0.7.3
 - Issue closed. External weights from averages saved in session (#11).
 - Solved minor bugs
 - Added estimation of errors of fit (Issue #12/#14)
 - Fixed: Some .fcs files containing averages were not imported.
0.7.2
 - Bugfix: Issue #10; we now have a goodness of the fit, if weighted
    fitting is performed
 - Bugfix: Weights for fitting not properly calculated (sqrt(std)).
 - Bugfix: Batch control IndexError with Info window opened
 - Tool Statistics: Sort values according to page numbers
 - Tool global: Added weighted fitting
 - Residuals: According to weighted fitting, weighted residuals are plotted
 - Average: Variances from averaging can be used for weighted fitting
0.7.1
 - Feature: Added Python shell
 - Bugfix: Saving image was not performed using WXAgg
 - Bugfix: Notebook pages were drag'n'dropable
 - Update function now works in its own thread
 - Code cleanup: documentation of model functions
 - Added program icon
0.7.0
 - File import dialog was enhanced (#4, #5 - subsequently #7, #8):
   - Now there is only one "load data" dialog in the file menu.
   - The model function is chosen for each type of data that is 
      to be imported (AC, CC, etc.).
   - Loading files that do not contain data a pointed out to the
     user and the program continues with the other files.
 - Bugfix: Channel selection window causes crash on file import (#1).
 - Bugfix: Hidden feature changes fixed parameters during fitting (#2).
 - Feature: Convert TIR model function parameters lambda and NA to sigma (#3).
 - Code cleanup: Opening data files is now handled internally differently.
0.6.9
 - Initital GitHub commit
