CHANGE.LOG

Change Log from 0.8.0 to 0.8.1
==============================

1) Bug fixes: 
  a) Version 0.8.0 was Python version sensitive. It behaved
     differently when run under versions before 2.5 and those
     at or after 2.5 due to changes in the way __import__
     worked.
  b) BUGLIST was not being update which each new release.

Change Log from 0.7.9 to 0.8.0
==============================

Converted distribution to use the standard Python setup.py method.
Thanks to Carlos Garcia Campos (carlosgc-at-gnome.org).

Change Log from 0.7.6 to 0.7.9
==============================

1) Bug fixes:
   a) ProcessRun was not executed under all circumstances.
   b) Code blocks in __main__ were ignored.
   c) Missing input files raised an exception and terminate after printing help.
   d) Class doc strings were counted as function doc strings.
   e) Metric module with missing metric class raised a 'key' exception.
   f) File with trailing whitespace on last indented line caused exception.
2) Output sent to stderr also sent to stdout. This allows user to see error
   in content since the output from stderr and stdout are not in sync.
3) Changed default metric class name so that for metric module wxYz,
   the default metric class name is WxYzMetric (note capitalized name)
   when the metric class name is omitted in the --includeMetrics
   parameter.
4) Added Decorators to newFeatures24 metric.


Change Log from 0.7.1 to 0.7.6
==============================
1) Added Features
   a) New option -B to suppress Basic Metrics output
   b) Added the new -e option. When the option is used,
      the SQL command file does not try to recreate
      the SQL tables. In other words, only INSERT
      commands are generated. Use this option if you are
      sure the SQL data and token tables already exist.
      This option implies the -N option (it deletes the
      old SQL command file, if it exists).
   c) Added computation of COCOMO 2's Source Lines Of Code
      (SLOC) metric. In COCOMO 2, SLOC is defined as:

        * Only Source lines that are delivered as part
          of the product are included.
        * Test drivers and other support software is excluded
        * Source lines are created by the project staff.
          Code created by applications generators is excluded
        * One SLOC is one logical line of code
        * Declarations are counted as SLOC
        * Comments are not counted as SLOC
        * Blank and empty lines are not counted as SLOC

      The COCOMO 1 model's SLOC is now called DSI (Delivered
      Source Instructions) in COCOMO 2.

      For example, in COCOMO 2, a Python "if-else" statement is
      counted as one SLOC, but might be counted as several DSIs.

      Given all that, PyMetrics will count test code if requested.
      I believe that this is correct to count test code because
      it is the basis of good Extreme Programming methodology.

      This SLOC metric also introduced raw text metrics. This
      has lead to a new member function of MetricBase:

        processSrcLines( self, srcLines, *a, **kw)

      where srcLines   is a string containing the raw text
                       of the current file.

      This member function is called once per file.

      The SLOC metric is computed by default.

      The following definitions are used to compute SLOC:

        1)  All literals are treated as occupying one logical
            line, regardless of the number of physical lines
            that are spanned.
        2)  All parenthesized lists, expressions, and argument
            lists are treated as occupying one logical line,
            regardless of the number of physical lines spanned.
        3)  All documentation strings are ignored.

2) Bug Fixes
   a) In Help, the default value displayed for --quiet
      was reversed. It now reads more logically.

3) Miscellaneous
   a) Now clearly identify generated code with a comment
      in the first line starting with

        "-- Automatically generated ..."

4) Added testSLOC.py to sample/ subdirectory. It tests most of
   weird cases that may occur in a Python source file due to
   continuations, literals of various types, and code spanning
   multiple lines. The source code shows what is counted with
   comments on the end of each counted source line. As of thie
   release, the SLOC metric for this file is 28.

Change Log from 0.6.0 to 0.7.1
==============================
Quite a lot has changed from the last version. The changes
fall into two categories:

1) Added Features
   a) More command line options to add flexibility to what
      is computed and displayed.

      PLEASE RUN PyMetrics TO DISPLAY ALL THE OPTIONS.

   b) Easier creation and inclusion of your own metrics
   c) Creation of a SQL table, called metricData, in the the
      metrics database that contains all the metrics that
      have been calculated. This table also tracks metrics
      over time, so you can compare metrics from one run
      to the next.
   d) Added library name field to SQL output. This is an
      user-defined name to identify a collection of modules.
      For example, all the Zope modules might have a 'zope'
      library name while all your modules might have your
      initials as a library name. In this way, you could
      compare the metrics in your library against the
      metrics in the Zope library.

      You could also use this feature to define a standard
      set of metrics with a library name of 'std' and then
      you could measure all other modules against the 'std'
      modules.

2) Refactoring and Reworking
   a) Due to the major enhancements introduced in this
      version, a number of modules have been refactored
      to simplify them and make them more testable.
   b) Unit tests have been written for a number of modules.
      The tests are contained in the "test" subdirectory.
      The "test" subdirectory has been added to this
      distribution.
3) Added missing LICENSE and "examples" subdirectory.
4) Fixed various bugs, none critical.

Change Log from 0.5.3 to 0.6
============================
1)  Modified PyMetrics to handle IOErrors more
    gracefully and continue executing, if
    possible.
2)  Check for lines ending in '\r' and convert character
    to blank since Python normally converts '\r' to '\n'
    on Windows and Mac systems.
3)  Changed metric format so each metric number appears
    before its name.
4)  Added the -f/--files option to allow the user to
    specify a file containing a list of modules to analyse.
5)  Fixed Halstead metric code to handle degenerate
    programs better.
6)  Changed SQL and CSV output to have symbolic names for
    types of names in fully-qualified function and class
    names.
7)  Changed semantic type of 'self' from VARNAME to KEYWORD.
8)  Now count any token of semantic significance, like
    keywords, function names, etc.
9)  Fixed issue where tarball dir was not created at a higher
    level directory with the name of the version in the tarball.
10) Counts of Token Types now listed in alphabetic order.
11) Added -K option to suppress display of token types.
12) Introduced MetricBase class for use in designing your
    own metrics. Member functions include:

    processToken( self, fcnName, className, tok, *args, **kwds )
    processStmt( self, fcnName, className, stmt, *args, **kwds )
    processBlock( self, fcnName, className, block, *args, **kwds )
    processFunction( self, fcnName, className, fcn, *args, **kwds )
    processClass( self, fcnName, className, cls, *args, **kwds )
    processModule( self, moduleName, module, *args, **kwds )
    processRun( self, run, *args, **kwds )

   where:

    context is context in which last token occurred.
    metrics is metrics up to the point where function called
    tok     is last completed token
    stmt    is last completed statement
    block   is last completed block
    fcn is last completed function
    cls is last completed class
    mod is last completed module
    run is last completed run (i.e., everything)

   Each of the stmt, block, fcn, cls, mod, and run is a
   list of tokens.

   The context is a dictionary related to the last token
   processed before invoking one of the processX member
   function. The context contains:

    inFile      current file name
    fcnDepth    current nested function depth
    classDepth  current nested class depth
    parenDepth  current parenthesis depth
    bracketDepth    current depth of brackets ([])
    braceDepth  current depth of braces ({})
    blockNum    current unique block number
    blockDepth  current nested block depth

13) By default, the Halstead metric is not displayed. It is still
    available, as explained by the Help. This makes the -H
    option redundant at this point. I have removed this metric
    because I have not found it to be generally useful. You may
    differ and include it in any of you runs.
14) By default, the token and keyword counts are not displayed.
    They are still available using the -k option. The -K option
    is still present, but redundant at this point. Again, I did
    not find this metric of particular use, so I made it
    optional.
15) By far, the most time spend in running PyMetrics is in
    writing out the SQL and CSV files. Suppressing one or both,
    if you do not need them, will speed up processing. However,
    I did speed up the processing of these files by 20-50%.


Change Log from 0.5.2 to 0.5.3
==============================
1) PyMetrics is now available under CVS.
2) Fixed a bug where the tokenize library module
   allowed the position of a comment on a line by
   itself to affect the indentation of subsequent
   blocks. This would cause a crash if the affected
   code included a function return statement.
3) Fixed bug where the McCabe metric routine was
   being called twice for most tokens.
4) Changed tokenize module to yield 5 parameters,
   as it originally did. At an early stage, I
   erroneously thought that I needed to pass back
   an extra value.
5) Sped up processing slightly by skipping redundant
   tests.
6) Changed subdirectory name from tests to examples.

Change Log from 0.5.1 to 0.5.2
==============================
This is a bug fix release.

1) Fixed bug where program failed if first line of
   file started with #!.

Change Log from 0.5.0 to 0.5.1
==============================
This is mainly a bug fix release. The changes made
also include:

1) Fixed several divide by zero bugs that could
   occur for degenerate Python programs.
2) Corrected version.py to reflect the correct
   version and name of the program.
3) Removed %DocStrings metric since it was
   meaningless.
4) Fixed bug where nested classes and nested
   functions were not correctly reported.
5) Changed format of report on classes/functions
   having or missing Doc strings.
6) Report on classes/functions now uses fully
   qualified names.
7) Refactored some code to common up handling of
   certain types of tokens.
8) Added the -z/--zero command line option. Normally,
   zero and empty metrics are not shown. This option
   forces the values to be displayed.

Change Log from 0.4.0 to 0.5.0
==============================
1) Simplified processing of command line arguents and
   parameter handling by objectifying parameter parsing.
2) Eliminated need for global variables when handling
   command line parameters and options.
3) Rearranged basic metrics to appear before computed
   metrics.
4) Moved code out of the main module, PyMetrics.py,
   into separate files.
5) Clarified some options in the Help output. Also
   added the -S option to suppress output of the SQL
   command file.
6) Defined and implemented endOfToken, endOfStmt,
   endOfFcn, endOfClass, endOfModule, and endOfRun
   functions for main program.
7) Implemented a very simple subscribe feature to
   allow various metric classes to be registered for
   the events in point 6).