IPYTHON AND CLEPY HAVE VERY LITTLE IN COMMON

Matt Wilson

CLEPY 2008


IPYTHON

Supports tab-completion:

    In [2]: mu.a
    mu.as_xml      mu.attrsearch

You can (usually) see the source for an object like this:

    In [2]: mu.attrsearch??

    Type:           function
    Base Class:     <type 'function'>
    String Form:    <function attrsearch at 0x83da87c>
    Namespace:      Interactive
    File:           /home/matt/lib/python/matt_utils.py
    Definition:     mu.attrsearch(o, attr_substr, r=None)
    Source:
    def attrsearch(o, attr_substr, r=None):

        """
        Return a list of any attributes of object o that have attr_substr in
        the name.
        o is any object that supports dir.
        attr_substr is a string.
        """
        if not r: return [a for a in dir(o) if attr_substr.lower() in a.lower()]
        else: return [a for a in dir(o) if attr_substr.lower() in a.lower()]

Every expression you type is stored in a dictionary In and the results
are stored in Out:

    In [20]: In[8]
    Out[20]: u'1 + 1\n'

    In [21]: Out[8]
    Out[21]: 2

The %ed magic command has some neat tricks:

    %ed # opens a temporary file.
    %ed ~/foo.py # opens or creates foo.py
    %ed mu.attrsearch  # opens the file where mu is defined in $EDITOR at the right line.
    %ed -p # opens the last file you just opened.
    %ed 8:10 14 # opens a new temporary file with what you put in on lines 8 through 10 and 14.

doctest mode is helpful:

    In [42]: chunkify??
    Type:           function
    Base Class:     <type 'function'>
    String Form:    <function chunkify at 0x8508374>
    Namespace:      Interactive
    File:           /home/matt/lib/python/matt_utils.py
    Definition:     chunkify(s, chunksize)
    Source:
    def chunkify(s, chunksize):
        "Yield sequence s in chunks of size chunksize."
        for i in range(0, len(s), chunksize):
            yield s[i:i+chunksize]

    In [43]: %doctest_mode
    Exception reporting mode: Plain
    Doctest mode is: ON

    >>> import matt_utils as mu

    >>> list(mu.chunkify('abcdefghi', 3))
    ['abc', 'def', 'ghi']

    >>> %doctest_mode
    Exception reporting mode: Context
    Doctest mode is: OFF

    In [50]: %ed mu.chunkify # paste in the docstring.

    In [51]: reload(mu)
    Out[51]: <module 'matt_utils' from '/home/matt/lib/python/matt_utils.py'>

    In [52]: mu.chunkify??
    def chunkify(s, chunksize):
        """
        Yield sequence s in chunks of size chunksize.

        >>> import matt_utils as mu

        >>> list(mu.chunkify('abcdefghij', 3))
        ['abc', 'def', 'ghi', 'j']

        """
        for i in range(0, len(s), chunksize):
            yield s[i:i+chunksize]

    In [65]: doctest.testmod(mu)

    ### SNIPPED OUT LOTS OF NOISE ####
    
    Out[65]: (0, 5)


Also, the pdb integration is really nice:

    In [67]: %pdb
    Automatic pdb calling has been turned ON

    In [68]: mu.chunkify(66, 2)
    Out[68]: <generator object at 0x84f74cc>

    In [69]: list(_68)
    ---------------------------------------------------------------------------
    TypeError                                 Traceback (most recent call last)

    /home/matt/svn-checkouts/<ipython console> in <module>()

    /home/matt/lib/python/matt_utils.py in chunkify(s, chunksize)
        140
        141     """
    --> 142     for i in range(0, len(s), chunksize):
        143         yield s[i:i+chunksize]
        144

    TypeError: object of type 'int' has no len()
    > /home/matt/lib/python/matt_utils.py(142)chunkify()
        141     """
    --> 142     for i in range(0, len(s), chunksize):
        143         yield s[i:i+chunksize]

    ipdb> l
    137
    138     >>> list(mu.chunkify('abcdefghij', 3))
    139     ['abc', 'def', 'ghi', 'j']
    140
    141     """
--> 142     for i in range(0, len(s), chunksize):
    143         yield s[i:i+chunksize]
    144
    145
    146 def as_xml(y, requestid=None, is_the_parent=True):
    147

    ipdb> a
    s = 66
    chunksize = 2
    ipdb>


And if you forgot to turn on %pdb in advance, you can also use %debug
after the fact.

    ipdb> q

    In [70]: %debug
    > /home/matt/lib/python/matt_utils.py(142)chunkify()
        141     """
    --> 142     for i in range(0, len(s), chunksize):
        143         yield s[i:i+chunksize]

    ipdb>

If you do this, then you'll always get an IPython session:

    $ cat ~/.pythonstartup
    import IPython
    IPython.Shell.IPShell().mainloop(sys_exit=1)

Finally, there's lots more cool stuff in IPython.  And IPython1
apparently is gonna be more like emacs SLIME where the editor and the
python interpreter talk to each other over the network.  So, you can
attach to a multi-threaded app while it is running and tinker with it.