Velocity Reviews > please help with optimisation of this code - update of given table according to another table

# please help with optimisation of this code - update of given table according to another table

Farraige
Guest
Posts: n/a

 11-08-2006
I am implementing the method that updates given table (table is
represented as list of lists of strings) according to other table (some
kind of merging)...

This method takes following arguments:
t1 - table we would like to update
t2 - table we would like to take data
from
keyColumns - list of key indexes e.g. [0,1]
columnsToBeUpdated - list of column indexes we would like to update in
our table T1 e.g [2,4]

Let's say we have a table T1:

A B C D E
---------------
1 4 5 7 7
3 4 0 0 0

and we call a method mergeTable(T1, T2, [0,1], [2,4])

It means that we would like to update columns C and E of table T1 with
data from table T2 but only in case the key columns A and B are equal
in both tables.... I grant that the given key is unique in both tables
so if I find a row with the same key in table T2 I do merging, stop and
go to next row in table T1...

Let's say T2 looks following:

A B C D E
---------------
2 2 8 8 8
1 4 9 9 9

So after execution of our mergeTable method, the table T1 should look
like :

A B C D E
1 4 9 7 9
3 4 0 0 0

The 2nd row ['3', '4', '0' ,'0', '0'] didn't change because there was
no row in table T2 with key = 3 ,4

The main part of my algorithm now looks something like ...

merge(t1, t2, keyColumns, columnsToBeUpdated)

........

for row_t1 in t1:
for row_t2 in t2:
if [row_t1[i] for i in keyColumns] == [row_t2[j] for j
in keyColumns]:
# the keys are the same
for colName in columnsToBeUpdated:
row_t1[colName] = row_t2[colName]

# go outside the inner loop - we found a row with
# the same key in the table
break

In my algorithm I have 2 for loops and I have no idea how to optimise
it (maybe with map? )
I call this method for very large data and the performance is a
critical issue for me

I will be grateful for any ideas

Marc 'BlackJack' Rintsch
Guest
Posts: n/a

 11-08-2006
In <(E-Mail Removed). com>, Farraige wrote:

> Let's say we have a table T1:
>
> A B C D E
> ---------------
> 1 4 5 7 7
> 3 4 0 0 0
>
> and we call a method mergeTable(T1, T2, [0,1], [2,4])
>
> It means that we would like to update columns C and E of table T1 with
> data from table T2 but only in case the key columns A and B are equal
> in both tables.... I grant that the given key is unique in both tables
> so if I find a row with the same key in table T2 I do merging, stop and
> go to next row in table T1...
>
> Let's say T2 looks following:
>
> A B C D E
> ---------------
> 2 2 8 8 8
> 1 4 9 9 9
>
> So after execution of our mergeTable method, the table T1 should look
> like :
>
> A B C D E
> 1 4 9 7 9
> 3 4 0 0 0
>
> The 2nd row ['3', '4', '0' ,'0', '0'] didn't change because there was
> no row in table T2 with key = 3 ,4
>
> The main part of my algorithm now looks something like ...
>
> merge(t1, t2, keyColumns, columnsToBeUpdated)
>
> .......
>
> for row_t1 in t1:
> for row_t2 in t2:
> if [row_t1[i] for i in keyColumns] == [row_t2[j] for j
> in keyColumns]:
> # the keys are the same
> for colName in columnsToBeUpdated:
> row_t1[colName] = row_t2[colName]
>
> # go outside the inner loop - we found a row with
> # the same key in the table
> break
>
> In my algorithm I have 2 for loops and I have no idea how to optimise
> it (maybe with map? )
> I call this method for very large data and the performance is a
> critical issue for me

Just go through the first table once and build a mapping key->row and then
go through the second table once and look for each row if the key is in
the mapping. If yes: update columns. This runs in O(2*rows) instead if
O(rows**2).

def update_table(table_a, table_b, key_columns, columns_to_be_updated):
def get_key(row):
return tuple(row[x] for x in key_columns)

key2row = dict((get_key(row), row) for row in table_a)
for row in table_b:
row_to_be_updated = key2row.get(get_key(row))
if row_to_be_updated is not None:
for column in columns_to_be_updated:
row_to_be_updated[column] = row[column]

def main():
table_a = [[1, 4, 5, 7, 7],
[3, 4, 0, 0, 0]]
table_b = [[2, 2, 8, 8, 8],
[1, 4, 9, 9, 9]]
update_table(table_a, table_b, (0, 1), (2, 4))
for row in table_a:
print row

Ciao,
Marc 'BlackJack' Rintsch

Antoon Pardon
Guest
Posts: n/a

 11-08-2006
On 2006-11-08, Farraige <(E-Mail Removed)> wrote:
>
> ...
>
> The main part of my algorithm now looks something like ...
>
> merge(t1, t2, keyColumns, columnsToBeUpdated)
>
> .......
>
> for row_t1 in t1:
> for row_t2 in t2:
> if [row_t1[i] for i in keyColumns] == [row_t2[j] for j
> in keyColumns]:
> # the keys are the same
> for colName in columnsToBeUpdated:
> row_t1[colName] = row_t2[colName]
>
> # go outside the inner loop - we found a row with
> # the same key in the table
> break
>
> In my algorithm I have 2 for loops and I have no idea how to optimise
> it (maybe with map? )
> I call this method for very large data and the performance is a
> critical issue for me
>
> I will be grateful for any ideas

One idea would be to precompute the list comprehensions in the if test.

p2 = [[row_t2[i] for i in keyColums] for row_t2 in t2]
for row_t1 in t1:
proj1 = [row_t1[i] for i in keyColumns]
for row_t1, proj2 in izip(t2, p2):
if proj1 == proj2:
...

--
Antoon Pardon

Gabriel Genellina
Guest
Posts: n/a

 11-08-2006
At Wednesday 8/11/2006 07:18, Farraige wrote:

> for row_t1 in t1:
> for row_t2 in t2:
> if [row_t1[i] for i in keyColumns] == [row_t2[j] for j
>in keyColumns]:
> # the keys are the same
> for colName in columnsToBeUpdated:
> row_t1[colName] = row_t2[colName]
>
> # go outside the inner loop - we found a row with
> # the same key in the table
> break
>
>In my algorithm I have 2 for loops and I have no idea how to optimise
>it (maybe with map? )
>I call this method for very large data and the performance is a
>critical issue for me

def getkey(row):
return tuple([row[i] for i in keyColumns])

Sort both tables using .sort(key=getkey); then traverse both in
paralell like in classic merge; in 1 pass you get the job done (not
counting the two initial sorts)

--
Gabriel Genellina
Softlab SRL

__________________________________________________
Correo Yahoo!
Espacio para todos tus mensajes, antivirus y antispam ˇgratis!
ˇAbrí tu cuenta ya! - http://correo.yahoo.com.ar

Farraige
Guest
Posts: n/a

 11-08-2006
Thank you all for your very valuable clues!
Thanks to you I got nearly 97% perfomance improvement !!!!!!!!!!!!!!! I
can't believe it ))
Once again thank you

Best wishes
Farraige