[Tutor] Starting to write a scanf-like module for Python

[Danny Yoo]

On Sun, 15 Aug 2004, Alan Gauld wrote:

> > I thought it might make a nice weekend project to write scanf for
> > Python; it's nowhere near done yet, but it's sorta working... *grin*
> >
> >     http://hkn.eecs.berkeley.edu/~dyoo/python/scanf.py
> >
> > I want to gauge some input from folks on the list to see if this might
> > be useful for folks.
>
> If it were sscanf() it would be really useful.

[some text cut]

> I would seriously consider sscanf() as a better project - and easier -
> and leave the acquiring the input string to raw_input or file.read()


Hi Alan,

Too late: I had already implemented it.  *grin*


In fact, I use it for my unit test cases.  Here's a sample of what I test:

###
def ScanfTests(unittest.TestCase):
    def testIntegerScanning(self):
        self.assertEquals((42, 43),
                          sscanf("   42\n   43  ", "%d %d"))

    def testWordScanning(self):
        self.assertEquals(("hello", "world"),
                          sscanf("   hello world", "%s %s"))

    def testSuppression(self):
        self.assertEquals((), sscanf(" hello world", "%*s %*s"))
###


I do know I need to document the module better, so I'll add some more
documentation strings.  What this 'scanf' module provides are the
following functions:

    scanf(formatString) -- reads from standard input
    sscanf(inputString, formatString) -- reads from string input
    fscanf(inputFile, formatString) -- reads from a file-like object


I've also mimicked the 're' module sligthly in the sense that the module
allows one to "compile" a format string, and then use the compiled pattern
over and over:

###
>>> import scanf
>>> patternReader = scanf.compile(" (%d, %d) ")
>>> patternReader
<function f at 0x850b0>
>>>
>>>
>>> from StringIO import StringIO
>>> sampleFile = StringIO('''
... (3, 3)
... (17, 42)
... (-1, 5)
... ''')
>>> sampleBuffer = scanf.CharacterBufferFromFile(sampleFile)
>>> patternReader(sampleBuffer)
(3, 3)
>>> patternReader(sampleBuffer)
(17, 42)
>>> patternReader(sampleBuffer)
(-1, 5)
>>> patternReader(sampleBuffer)
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
  File "scanf.py", line 310, in f
    raise IncompleteCaptureError, (e, tuple(results))
scanf.IncompleteCaptureError: (<scanf.FormatError instance at 0x84580>,
())
###

Hmmm... I guess I should simplify the usage here so that the user doesn't
have to manually call the internal "CharacterBufferFromFile" thingy.



> scanf() has so many bugs/features nobody in their right mind uses it do
> they?

Yes, that's the problem.  So I'm not quite sure what features people
really find useful in scanf().  *grin*

But I personally need something that does integer and word scanning, so
that's what I've implemented so far.  I don't mind if this does only a
subset of C's scanf(), just as long as it does the most important stuff,
the stuff that people actually use with the *scanf() C functions.