C[omp]ute

Welcome to my blog, which was once a mailing list of the same name and is still generated by mail. Please reply via the "comment" links.

Always interested in offers/projects/new ideas. Eclectic experience in fields like: numerical computing; Python web; Java enterprise; functional languages; GPGPU; SQL databases; etc. Based in Santiago, Chile; telecommute worldwide. CV; email.

Personal Projects

Choochoo Training Diary

Last 100 entries

Extended Date Parsing in Python

From: andrew cooke <andrew@...>

Date: Sun, 23 Jun 2013 04:40:29 -0400

I've modified the way that Python parses dates to make it significantly more
flexible.  As a consequence, SimpleDate can now parse dates mure more
quickly.  For example, all these dates are parsed by a single template (before
they would have required evaluating different templates in turn):

    >>> SimpleDate('2013-06-23').normalized
    SimpleDate('2013-06-23 04:00:00.000000 UTC', tz='UTC')
    >>> SimpleDate('20130623T0420').normalized
    SimpleDate('2013-06-23 08:20:00.000000 UTC', tz='UTC')
    >>> SimpleDate('2013/06/23 4:20 UTC').normalized
    SimpleDate('2013-06-23 04:20:00.000000 UTC', tz='UTC')

(the 4 hours shift on parsing the first two is because by default we use the
locale timezone, which is CLT).


The extension to Python's date templates is conservative in that:

 - existing symbols have the same meaning as before

 - the construction of a date from the parsed data uses exactly the same
   algorithm as before


What has been added is the ability to:

 - Specify alternatives.  For example, {%Z|%z} will match either of the
   two timezone formats.

 - Specify optional values.  For example %Y-%m-%d{ %H:%M:%S}? makes the
   time part optional.


As you can see from the examples, the extension uses familiar regexp syntax,
but with brackets instead of parentheses (parens sometimes occur in date
expressions, so this reduces the need for escaping).


What's particularly cute about the implementation is that parsing will
"disambiguate" the template.  You get back from the parser an equivalent
template that matches (more or less) the parsed input, without the extra
features.  This allows you to use display the date in the original format.

For example:

    >>> strptime('12:34', '{%H:}?%M')
    ((1900, 1, 1, 12, 34, 0, 0, 1, -1, None, None), 0, '%H:%M')
    >>> strptime('34', '{%H:}?%M')
    ((1900, 1, 1, 0, 34, 0, 0, 1, -1, None, None), 0, '%M')

in the second case no hours were given and the resturned format (final value
in result) is only for the minutes.


Simple Date is available at:
  https://github.com/andrewcooke/simple-date
  https://pypi.python.org/pypi/simple-date

The source that implements this can be seen here:
  https://github.com/andrewcooke/simple-date/blob/master/src/simpledate/fmt.py

Andrew

Bad Design

From: andrew cooke <andrew@...>

Date: Sun, 23 Jun 2013 18:34:12 -0400

So, that wasn't so smart.

The problem is that the transformation from extended to normal date templates
is not idempotent.  In practice, that means that if you have something like

   {%H}

you have no way of knowing if that is a valid write format (which prints the
hour in brackets) or a valid read format (that creates a pointless group
around the hour, matching nothing more than the digits).


The only reasonable solution is to change the syntax:

 - alternates and grouping look like  %(...%|...%|...%)
   (may as well use parens since it's escaped anyway).

 - modified values look like  %!Z  (the ! comes "in front")


I will probably implement this tonight.

Andrew

Comment on this post