1. Armin Rigo
  2. cpython-withatomic


cpython-withatomic / Mac / Relnotes-1.1

The branch 'legacy-trunk' does not exist.


For the most part, Python on the Mac works just like Python under UNIX.
The most important differences are:

- Since there is no shell environment on the Mac, the start-up file
  has a fixed name: PythonStartup.  If a file by this name exists
  (either in the current folder or in the system folder) it is executed
  when an interactive interpreter is started.

- The default search path for modules is different: first the current
  directory is searched, then the subdirectories 'lib', 'lib:stdwin' and
  'demo'.  As always, you can change this (e.g. in your PythonStartup
  file) by assigning or appending to sys.path -- use Macintosh pathnames!
  (The default contains no absolute paths because these are unlikely
  to make sense on other people's hard disks.)

- The user interface for typing interactive commands is different.
  This is actually the THINK C console I/O module, which is based on
  the Mac toolbox TextEdit.  A standard Edit menu provides Cut, Copy,
  Paste and Clear (Undo is only there for Desk Accessories).  A minimal
  File menu provides Quit, which immediately exits the application,
  without the usual cleanup.  You can Copy from previous output,
  but you can't scroll back beyond the 24x80 screen.  The TAB key
  always brings you to the end of the current input line; indentation
  must be entered with spaces (a single space is enough).
  End-of-file is generated by Command-D; Command-Period interrupts.
  There is an annoying limit in the length of an input line to a single
  screen line (less the prompt).  Use \ to input long statements.
  Change your program if it requires long lines typed on input.
  Even though there is no resize box, the window can be resized by
  dragging its bottom right corner, but the maximum size is 24x80.

- Tabs in module files are interpreted as 4 (four!) spaces.  This is
  consistent with most Mac editors that I know.  For individual files
  you can change the tab size with a comment like

	# vi:set tabsize=8:

  (exactly as shown here, including the colons!).  If you are consistent
  in always using tabs for indentation on UNIX, your files will be
  parsed correctly on the Mac, although they may look funny if you
  have nicely lined-up comments or tables using tabs.  Never using tabs
  also works.  Mixing tabs and spaces to simulate 4-character indentation
  levels is likely to fail.

- You can start a script from the Finder by selecting the script and
  the Python interpreter together and then double clicking.  If you
  make the owner of the script PYTH (the type should always be TEXT)
  Python will be launched if you double click it!
  There is no way to pass command line arguments to Python scripts.

- The set of built-in modules is different:

  = Operating system functions for the 'os' module is provided by the
    built-in module 'mac', not 'posix'.  This doesn't have all the
    functions from posix, for obvious reasons (if you know the Mac
    O/S a little bit).  The functions in os.path are provided by
    macpath, they know about Mac pathnames etc.
  = None of the UNIX specific modules ('socket', 'pwd', 'grp' etc.)
  = Module 'stdwin' is always available.  It uses the Mac version of
    STDWIN, which interfaces directly with the Mac toolbox.  The most
    important difference is in the font names; setfont() has a second
    argument specifying the point size and an optional third one
    specifying the variation: a single letter character string,
    'i' for italics, 'b' for bold.  Note that when STDWIN is waiting
    for events, the standard File and Edit menus are inactive but
    still visible, and (most annoyingly) the Apple menu is also inactive;
    conversely, menus put up by STDWIN are not active when the Python is
    reading from the keyboard.  If you open Python together with a text
    file containing a Python script, the script will be executed and
    a console window is only generated when the script uses standard
    input or output.  A script that uses STDWIN exclusively for its I/O
    will have a working Apple menu and no extraneous File/Edit menus.
    (This is because both stdwin and stdio try to initialize the
    windowing environment; whoever gets there first owns the Apple menu.)
    LIMITATIONS: a few recent additions to STDWIN for X11 have not yet
    been added to the Mac version.  There are no bitmap objects, and
    the setwinpos() and setwinsize() methods are non--functional.

- Because launching an application on the Mac is so tedious, you will
  want to edit your program with a desk accessory editor (e.g., Sigma
  edit) and test the changed version without leaving Python.  This is
  possible but requires some care.  Make sure the program is a module
  file (filename must be a Python identifier followed by '.py').  You
  can then import it when you test it for the first time.  There are
  now three possibilities: it contains a syntax error; it gets a runtime
  error (unhandled exception); or it runs OK but gives wrong results.
  (If it gives correct results, you are done testing and don't need
  to read the rest of this paragraph. :-)  Note that the following
  is not Mac-specific -- it's just that on UNIX it's easier to restart
  the entire script so it's rarely useful.
  Recovery from a syntax error is easy: edit the file and import it
  Recovery from wrong output is almost as easy: edit the file and,
  instead of importing it, call the function reload() with the module
  name as argument (e.g., if your module is called foo, type
  Recovery from an exception is trickier.  Once the syntax is correct,
  a 'module' entry is placed in an internal table, and following import
  statements will not re-read the file, even if the module's initialization
  terminated with an error (one reason why this is done is so that
  mutually recursive modules are initialized only once).  You must
  therefore force re-reading the module with reload(), however, if this
  happens the first time you try to import the module, the import statement
  itself has not completed, and your workspace does not know the module
  name (even though the internal table of moduesl does!).  The trick is
  to first import the module again, then reload it.  For instance,
  "import foo; reload(foo)".  Because the module object already exists
  internally, the import statement does not attempt to execute the
  module again -- it just places it in your workspace.
  When you edit a module you don't have to worry about the corresponding
  '.pyc' file (a "compiled" version of the module, which loads much faster
  than the textual version): the interpreter notices that the '.py' file
  has changed (because its modification time has changed) and ignores the
  '.pyc' file.  When parsing is successful, a new '.pyc' file is written;
  if this fails (no write permission, disk full or whatever) it is
  silently skipped but attempted again the next time the same module
  is loaded.  (Thus, if you plan to place a Python library on a read-only
  disk, it is advisable to "warm the cache" by making the disk writable
  and importing all modules once.  The standard module 'importall' helps
  in doing this.)