tips requested for a log-processing script

[Jaap]

Python ers,
As a relatively new user of Python I would like to ask your advice on
the following script I want to create.

I have a logfile which contains records. All records have the same
layout, and are stored in a CSV-format. Each record is (non-uniquely)
identified by a date and a itemID. Each itemID can occur 0 or more times
per month. The item contains a figure/amount which I need to sum per
month and per itemID. I have already managed to separate the individual
parts of each logfile-record by using the csv-module from Python 2.5.
very simple indeed.

Apart from this I have a configuration file, which contains the list of
itemID's i need to focus on per month. Not all itemID's are relevant for
each month, but for example only every second or third month. All
records in the logfile with other itemID's can be ignored. I have yet to
define the format of this configuration file, but am thinking about a 0
or 1 for each month, and then the itemID, like:
"1 0 0 1 0 0 1 0 0 1 0 0 123456" for a itemID 123456 which only needs
consideration at first month of each quarter.

My question to this forum is: which data structure would you propose?
The logfile is not very big (about 200k max, average 200k) so I assume I
can store in internal memory/list?

How would you propose I tackle the filtering of relevant/non-relevant
items from logfile? Would you propose I use a filter(func, list) for
this task or is another thing better?

In the end I want to mail the outcome of my process, but this seems
straitforward from the documentation I have found, although I must
connect to an external SMTP-server.

Any tips, views, advice is highly appreciated!
Jaap

PS: when I load the logfile in a spreadsheet I can create a pivot table
which does about the same ;-] but that is not what I want; the
processing must be automated in the end with a periodic script which
e-mails the summary of the keyfigure every month.

[martdi]

if you are running in windows you can use the win32com module to
automate the process of generating a pivot table in excel and then code
to send it via e-mail

Jaap wrote:

Python ers,
As a relatively new user of Python I would like to ask your advice on
the following script I want to create.

I have a logfile which contains records. All records have the same
layout, and are stored in a CSV-format. Each record is (non-uniquely)
identified by a date and a itemID. Each itemID can occur 0 or more times
per month. The item contains a figure/amount which I need to sum per
month and per itemID. I have already managed to separate the individual
parts of each logfile-record by using the csv-module from Python 2.5.
very simple indeed.

Apart from this I have a configuration file, which contains the list of
itemID's i need to focus on per month. Not all itemID's are relevant for
each month, but for example only every second or third month. All
records in the logfile with other itemID's can be ignored. I have yet to
define the format of this configuration file, but am thinking about a 0
or 1 for each month, and then the itemID, like:
"1 0 0 1 0 0 1 0 0 1 0 0 123456" for a itemID 123456 which only needs
consideration at first month of each quarter.

My question to this forum is: which data structure would you propose?
The logfile is not very big (about 200k max, average 200k) so I assume I
can store in internal memory/list?

How would you propose I tackle the filtering of relevant/non-relevant
items from logfile? Would you propose I use a filter(func, list) for
this task or is another thing better?

In the end I want to mail the outcome of my process, but this seems
straitforward from the documentation I have found, although I must
connect to an external SMTP-server.

Any tips, views, advice is highly appreciated!
Jaap

PS: when I load the logfile in a spreadsheet I can create a pivot table
which does about the same ;-] but that is not what I want; the
processing must be automated in the end with a periodic script which
e-mails the summary of the keyfigure every month.

[George Sakkis]

Jaap wrote:

Apart from this I have a configuration file, which contains the list of
itemID's i need to focus on per month. Not all itemID's are relevant for
each month, but for example only every second or third month. All
records in the logfile with other itemID's can be ignored. I have yet to
define the format of this configuration file, but am thinking about a 0
or 1 for each month, and then the itemID, like:
"1 0 0 1 0 0 1 0 0 1 0 0 123456" for a itemID 123456 which only needs
consideration at first month of each quarter.

It's probably not necessary if your records are in the order of 100K,
but if you're dealing with millions and above, you can write your
config file in binary using the struct module and condense it down to 6
bytes per record (32 bits for the ID and 12 bits for the months
occurences). Filtering will also be faster, as for each record you just
have to do a bitwise AND with the 0..010...0 mask corresponding to a
given month.

George

[Hendrik van Rooyen]

"Jaap" <ja**@nospaml.comwrote:

Python ers,
As a relatively new user of Python I would like to ask your advice on
the following script I want to create.

I have a logfile which contains records. All records have the same
layout, and are stored in a CSV-format. Each record is (non-uniquely)
identified by a date and a itemID. Each itemID can occur 0 or more times
per month. The item contains a figure/amount which I need to sum per
month and per itemID. I have already managed to separate the individual
parts of each logfile-record by using the csv-module from Python 2.5.
very simple indeed.

Apart from this I have a configuration file, which contains the list of
itemID's i need to focus on per month. Not all itemID's are relevant for
each month, but for example only every second or third month. All
records in the logfile with other itemID's can be ignored. I have yet to
define the format of this configuration file, but am thinking about a 0
or 1 for each month, and then the itemID, like:
"1 0 0 1 0 0 1 0 0 1 0 0 123456" for a itemID 123456 which only needs
consideration at first month of each quarter.

My question to this forum is: which data structure would you propose?
The logfile is not very big (about 200k max, average 200k) so I assume I
can store in internal memory/list?

How would you propose I tackle the filtering of relevant/non-relevant
items from logfile? Would you propose I use a filter(func, list) for
this task or is another thing better?

In the end I want to mail the outcome of my process, but this seems
straitforward from the documentation I have found, although I must
connect to an external SMTP-server.

Any tips, views, advice is highly appreciated!
Jaap

PS: when I load the logfile in a spreadsheet I can create a pivot table
which does about the same ;-] but that is not what I want; the
processing must be automated in the end with a periodic script which
e-mails the summary of the keyfigure every month.

I would do something like this: (obviously untested)

for line in readlines(open(logfile,r,1)):
(code to get hold of item, date, amount)
if item not in item_dict:
item_dict[item] = [(date,amount)]
else:
item_dict[item].append(date,amount)

this will give you, for each unique item, a direct ref to wherever its been
used.

I would then work through the config file, and extract the items of interest for
the run date...

HTH - Hendrik

[Jaap]

Hendrik van Rooyen schreef:

"Jaap" <ja**@nospaml.comwrote:

>Python ers,

Thanks!
all your replies have been both to the point and helpfull for me.

You have proven both Python and it's community are open and welcoming to
new users.

Jaap