This seems overbuilt, and would also be more tersely expressed with Numpy/Pandas. Additionally, your manual loops - whereas you're satisfied with current performance - will be faster in vectorised form.

This program can be as simple as

import numpy as npimport pandas as pdfrom string import ascii_uppercasefrom numpy.random import default_rnground_precision = 3nrows = 10rand = default_rng()array_3 = rand.choice(tuple(ascii_uppercase), size=(nrows, 5))df = pd.DataFrame({    'col1': rand.integers(low=1e4, high=1e5, size=nrows),    'col2': rand.uniform(low=1e4, high=1e5, size=nrows).round(decimals=round_precision),    'col3': pd.DataFrame(array_3).agg(''.join, axis=1),    'col4': rand.integers(low=10, high=16, size=nrows),    'col5': rand.uniform(low=5.5, high=6.7, size=nrows).round(decimals=round_precision),    'col6': rand.integers(low=1, high=6, size=nrows),    'col7': rand.choice(np.linspace(start=1.1, stop=3.3, num=3), size=nrows),    'col8': rand.choice(('A', 'AB'), size=nrows),})df.to_csv('eightVals.csv', index=None)

with output

col1,col2,col3,col4,col5,col6,col7,col890510,54337.357,TITJJ,12,6.254,2,2.2,A82827,10498.364,MLZFY,11,5.881,5,2.2,A81292,95873.98,UOHVL,14,5.931,2,1.1,AB44278,69859.781,RGQVO,11,6.492,2,2.2,AB26424,18106.356,XPCYV,15,6.176,5,2.2,AB12324,52143.779,VUEWA,15,5.959,2,3.3,A38774,34881.201,URXXL,15,6.545,2,2.2,AB25801,31699.204,XEWGS,14,6.427,3,3.3,A59555,98153.322,WYGBV,11,6.062,1,2.2,A44449,86569.519,MDNXN,15,6.229,1,3.3,A

If this needs to be reusable (and I'm not entirely convinced that it is, given what you've shown), you can adapt the above to utility functions. I don't know that classes are necessary.

• 1
  \$\begingroup\$Storing the config separately makes a lot of sense in data pipelines IMHO. Classes don't seem necessary just for that indeed though.\$\endgroup\$

  Lodinn

  – Lodinn

  2022-10-18 12:53:17 +00:00

  CommentedOct 18, 2022 at 12:53

I like how you started designing withgenerate method as interface for your generators. I like how you define your schema. I like also how the final loop and code works.

Really can't agree with the way you implemented yourgenerate methods. First it was incredibly confusing and it took me unnecessarily too long to figure out, what are these classes about. To be more specific:

• Why passing params as dict? Why not pass named parameters and make them instance variables? It would make code much more readable and easier to use. Only reason I can think about is that you could instantiate all those classes with same type of parameters (but you don't need to do that).
• That bigif in yourgenerate method is code smell. Each of those if branches should be it's own class with it's owngenerate implementation.
• NamesFixedLength andFromPossibleValues can be better.

Some possibilities examples how to implement your classes better (imho):

class FixedLengthIntGenerator(BoundType):    def __init__(self, length):        self.length = length    def generate(self):        length = self.length            return rnd.randint(10 ** (length - 1), 10 ** length - 1)# TODO: other classes for string, float, range ...class FromPresetChoicesGenerator(BoundType):    def __init__(self, choices):        self.choices = choices    def generate(self):       return rnd.choice(self.choices)

Yes it is more verbose, but it is more readable, extensible, different implementations are not so bound. Also it's not necessary to pass this "dType" anymore. Instead, you choose what implementation to use.

• 3
  \$\begingroup\$If you use a class having two methods, one of which is__init__, you should be using a function instead. Additionally using inheritance without a single call tosuper() or any of the parent class' features, is a big no no for me.\$\endgroup\$

  user51621

  – user51621

  2022-10-18 05:52:45 +00:00

  CommentedOct 18, 2022 at 5:52

• python
• python-3.x