The simplest design is for a command line program. The program will provide a help feature explaining its options. Once the command line program is built it would be relatively easy to add an input dialog GUI interface on top of it. This last step will occur as time permits. The changes to the original program are described below.
An example use of the program would be:
> python gpcp2csv.py -h
This could also be entered as:
> gpcp2csv.py --help
and provides all the options such as below:
usage: gpcp_to_csv.py [-h] [-z] [-s] [-i INPUT_FILE] [-o OUTPUT_FILE]
[-p INPUT_PREFIX] [-op OUTPUT_PREFIX] [-y YEARS]
[-m MONTHS] [-f FORMAT]
Please see http://precip.gsfc.nasa.gov/ for a full description of the data
set.
optional arguments:
-h, --help show this help message and exit
-z, --gzip Set this flag if the input files are zipped using gzip
and end with .gz.
-s, --single_file Set this flag if running the program for only a single
(one) file. Specify the file using the -i option.
-i INPUT_FILE, --input_file INPUT_FILE
The full path of the input file, i.e.:
C:/Users/schiefej/Desktop/gpcp_v2.2psg.1987 if no
year(s) is(are) provided this is required. If this
switch is used, -s must also be set.
-o OUTPUT_FILE, --output_file OUTPUT_FILE
The full path of the input file, i.e.:
C:/Users/schiefej/Desktop/out_1987.csv
-p INPUT_PREFIX, --input_prefix INPUT_PREFIX
The path to the input file(s) and the appropriate
prefix such as: './gpcp_v2.2_psg.' or
'C:/Users/schiefej/binary_files/gpcp_file_'The default
is the current directory plus the prefix for the new
format files: './gpcp_1dd_v1.2_p1d.'
-op OUTPUT_PREFIX, --output_prefix OUTPUT_PREFIX
The path to the output file(s) and the appropriate
prefix such as: './gpcp_out_' or
'C:/Users/schiefej/binary_files/gpcp_out_'
-y YEARS, --years YEARS
Years of the requested data, i.e. 2012 or 1987,2011.
The MUST be no spaces in the string, ie. 1999,2000 is
okay but 1999, 2000 is not.This is a optional value
but if it is provided, prefixes must be provided. If
it is provided, an input file must be provided.
-m MONTHS, --months MONTHS
Months of the requested data, i.e. 10 or
1,2,3,4,5,6,7,8,9,10,11,12 for an entire year. There
MUST be no spaces in the string; see the year argument
for applicable examples but using month numbers.
-f FORMAT, --format FORMAT
This switch determines which parser to use. 0 decodes
MONTHLY GPCP (v2.2) binary file into CSV output. Tags
the data by month and pre-pends a list of grid box
lat/lons in the same order as the precip data itself.
Format originally designed by Adler, et al.1 decodes
MONTHLY GPCP (v2.2) binary data into CSV with a
date/lat/lon value for each precip value in the file.
One line per value. 2 decodes DAILY GPCP (v1.2) binary
file into same csv format as 1 above.
Special thanks to clawtros/Adam Benzan and Carbon Chick for their initial
program that solved the hard problem.
Users of Unix/Linux environments will find the above familiar and easy to follow.
Added the get_variables() function. The dictionary returned by this function is used then to determine file dimensions, date and other bits of information needed in the remainder of the class.
"... premature optimization is the root of all evil." -- Donald Knuth
Given this I took some time to try some optimizations to get some practical experience. The only significant improvement to be made once a profiler was run was to use available memory and only write to the file once. This is where the cStringIO.StringIO() function is used and was admittedly only a marginal optimization at best but it was perceptible and was left in the code. If the data set increase in size significantly this may become problematic depending on the memory of the machine. For now it is a good trade off and readability of the code did not suffer.
The place for formatting the code is in the GpcpParser.write_csv function. This single function controls how the output is formatted. As stated in the "Notes" in Requirements, deriving from GpcpParser and then writing separate functions each format is one way of solving this problem.