FMTL: a Field-Mappable Tuple List

FMTL provides a basic way to make data transformations on Python 3(.6) tuple lists without pre-established dependencies. At its core, FMTL is plain simple (~100 loc). The idea is to ease dataset creation and replace standard data transformation boilerplate code.

Install:

via pip: pip install -e git+https://github.com/cedias/FMTL.git

Usage

Let's say you have a tuple list tl and that you want to make some transformations:

• Transform the str "id" field to a unique int identifier
• Multiply each "num" by 2

First create the FMTL object:

from fmtl import FMTL

tl = [("id0",1,44),("id0",5,75),("id1",3,44)] #raw tuple list
fmtl = FMTL(tl,{0:"id",1:"type",2:"num"}) #create the FMTL (rows can be a tuple or list)

Then you can easily add functional mappings to fields with the set_mapping(...) method:

#Multiply all "num" values by 2
fmtl.set_mapping("num",lambda x:x*2)

You can provide dictionnaries as mapping an the __get_item__() function will be automatically used.

fmtl.set_mapping("id", map)  #sets it

Mapping is done as iteration time in the object's __get_item__() method.

list(fmtl) ==> [(0, 1, 44), (0, 5, 75), (1, 3, 44)] 
list(fmtl.tuplelist) =>[('id0', 1, 44), ('id0', 5, 75), ('id1', 3, 44)]

You still can prebuild it explicitely, it's just a list after all.

#persist mapping explicitely.
builded = list(fmtl) #if you don't have too much data
builded = list(x for x in tqdm(fmtl)) #you can use tqdm to track progress if needed.
builded => [(0, 1, 44), (0, 5, 75), (1, 3, 44)]