Ryex - 4 months ago 14

Python Question

I'm using the following class to create a table and I need to find a way to not only make it faster but make interactions with it faster:

`class Table(object):`

"""a three dimensional table object"""

def __init__(self, xsize=1, ysize=1, zsize=1):

self.xsize = xsize

self.ysize = ysize

self.zsize = zsize

self.data = [0] * (xsize * ysize * zsize)

def __getitem__(self, key):

x, y, z = self.__extractIndices(key)

return self.data[x + self.xsize * (y + self.ysize * z)]

def __setitem__(self, key, value):

x, y, z = self.__extractIndices(key)

self.data[x + self.xsize * (y + self.ysize * z)] = value

def __extractIndices(self, key):

x = y = z = 0

if (self.ysize > 1):

if (self.zsize > 1):

if len(key) != 3:

raise IndexError

else:

x, y, z = key

elif len(key) != 2:

raise IndexError

else:

x, y = key

elif not isinstance(key, int):

raise IndexError

else:

x = key

return (x, y, z)

def resize(self, xsize=1, ysize=1, zsize=1):

"""resize the table preserving data"""

oldlist = list(self.data)

self.data = [0] * (xsize * ysize * zsize)

self.xsize = xsize

self.ysize = ysize

self.zsize = zsize

for i in range(0, oldlist):

self.data[1] = oldlist[i]

at on point I need to find if the data in two lists is equivalent of each of the z's

so I did this.

`self.data`

`self.map.data`

`for x in range(self.map.width - 1):`

for y in range(self.map.height - 1):

tempflag = False

#layer 1

if self.data[x, y, 0] != self.map.data[x, y, 0]:

tempflag = True

layer1flag = True

#layer 2

if self.data[x, y, 1] != self.map.data[x, y, 1]:

tempflag = True

layer2flag = True

#layer 3

if self.data[x, y, 2] != self.map.data[x, y, 2]:

tempflag = True

layer3flag = True

#copy the data if it changed

if tempflag:

self.data = copy.deepcopy(self.map.data)

previewflag = True

clearly this is the slowest way I could conceivably do this and considering that some of these tables I'm comparing have a size of 200 * 200 * 3 = 120,000 entries. I NEED this to be as fast as possible.

I've considered rewriting the above comparison to slice all the entries for one z like so

`tempflag = False`

#layer 1

slicepoint1 = 0

slicepoint2 = self.data.xsize * self.data.ysize * 1

data1 = self.data.data[slicepoint1:slicepoint2]

data2 = self.map.data.data[slicepoint1:slicepoint2]

if data1 != data2:

tempflag = True

layer1flag = True

#layer 2

slicepoint1 = self.data.xsize * self.data.ysize * 1

slicepoint2 = self.data.xsize * self.data.ysize * 2

data1 = self.data.data[slicepoint1:slicepoint2]

data2 = self.map.data.data[slicepoint1:slicepoint2]

if data1 != data2:

tempflag = True

layer2flag = True

#layer 3

slicepoint1 = self.data.xsize * self.data.ysize * 2

slicepoint2 = self.data.xsize * self.data.ysize * 3

data1 = self.data.data[slicepoint1:slicepoint2]

data2 = self.map.data.data[slicepoint1:slicepoint2]

if data1 != data2:

tempflag = True

layer3flag = True

#copy the data if it changed

if tempflag:

self.data = copy.deepcopy(self.map.data)

previewflag = True

and while this seems like it would go faster it still seems like it's could be significantly improved. for example could a not use numpy to build the data list inside the Table class?

I need this class and this check to run as fast as it possibly can

it would also be nice if the use of numpy would allow me to loop through the table really fast so I could use the data in it for blit operations to build a tilemap

I

Answer

This is definitely an application for NumPy! It will not only speed up your code, it will also simplify your code considerably, because indexing and comparison are already handled by NumPy. You will have to read some tutorial to learn NumPy -- just a few hints to get you going in this case.

Usually, I would simply derive from `numpy.ndarray`

to define a custom array class, but you stated that you definitely need the `data`

attribute, which clashes with `numpy.ndarray.data`

. Your class simplifies to

```
class Table(object):
def __init__(self, xsize=1, ysize=1, zsize=1):
self.data = numpy.zeros((xsize, ysize, zsize))
def __getitem__(self, key):
return self.data[key]
def __setitem__(self, key, value):
self.data[key] = value
def resize(self, xsize=1, ysize=1, zsize=1):
# This only works for increasing the size of the data,
# but is easy do adapt to other cases
newdata = numpy.zeros((xsize, ysize, zsize))
shape = self.data.shape
newdata[:shape[0], :shape[1], :shape[2]] = self.data
self.data = newdata
```

Your comparison code simplifies to

```
eq = self.data == self.data.map
layerflags = eq.reshape(-1, 3).any(axis=0)
if layerflags.any():
self.data[:] = self.map.data
```

And it will be **much** faster too!