cdxcore 0.1.71

Basic Python Tools; upgraded cdxbasics

cdxcore documentation

This module contains a number of lightweight tools, developed for managing data analytics and machine learning projects.

Install using:

pip install -U cdxcore

Documentation can be found here: https://quantitative-research.de/docs/cdxcore. cdxcore is best used with Python 3.12 and above, but is tested vs Python 3.10 onwards.

Highlights

• Dynamic plotting: simple live/animated plots built on Matplotlib.
• Config management: validated, discoverable configurations with automatic help.
• Versioning & caching: code-versioned I/O and reproducible hashing for pipelines.
• PrettyObject: dictionary-like objects that allow attribute access.
• Utilities: formatting helpers, binary I/O, shared-memory arrays, and more.

Main Functionality

• cdxcore.dynaplot is a framework for simple dynamic graphs with matplotlib. It has a simple methodology for animated updates for graphs (e.g. during training runs), and allows generation of plot layouts without knowing upfront the number of plots (e.g. for plotting a list of features).

  

• cdxcore.config allows robust management of configurations. It automates help, validation checking, and detects misspelled configuration arguments.

  from cdxcore.config import Config, Int, Float
  
  class Network(object):
      def __init__( self, config ):
          self.depth      = config("depth", 1, Int>0, "Depth of the network")
          self.width      = config("width", 1, Int>0, "Width of the network")
          self.activation = config("activation", "selu", str, "Activation function")
          config.done() # see below
  
  config = Config()
  config.network.depth         = 10
  config.network.width         = 100
  config.network.activation    = 'relu'
  
  network = Network(config.network)
  config.done()
  

• cdxcore.subdir wraps various file and directory functions into convenient objects. Useful if files have common extensions.

  from cdxcore.subdir import SubDir
  import numpy as np
  root   = SubDir("!")   # current temp directory
  subdir = root("test")  # sub-directory 'test'
  subdir.write("data", np.zeros((10,2)))
  data   = subdir.read("data")
  

• Caching: SubDir supports code-versioned file i/o which is used by @cdxcore.subdir.SubDir.cache for an efficient code-versioned caching protocol for functions and objects:

from cdxcore.subdir import SubDir
cache   = SubDir("!/.cache;*.bin")

@cache.cache("0.1")
def f(x,y):
   return x*y

_ = f(1,2)    # function gets computed and the result cached
_ = f(1,2)    # restore result from cache
_ = f(2,2)    # different parameters: compute and store result

• Code versioning is implemented in cdxcore.version:

  from cdxbasics.version import version
  
  @version("0.0.1")
  def f(x):
      return x
  
  print( f.version.full )   # -> 0.0.1
  

• Hashing (which is used for caching above) is implemented in cdxcore.uniquehash:

  class A(object):
      def __init__(self, x):
          self.x = x
          self._y = x*2  # protected member will not be hashed by default
  
  from cdxcore.uniquehash import UniqueHash
  uniqueHash = UniqueHash(length=12)
  a = A(2)
  print( uniqueHash(a) ) # --> "2d1dc3767730"
  

• cdxcore.pretty provides a PrettyObject class whose objects operate like dictionaries. This is for users who prefer attribute . notation over item access when building structured output.

  from cdxbasics.prettydict import PrettyObject
  pdct = PrettyObject(z=1)
  
  pdct.num_samples = 1000
  pdct.num_batches = 100
  pdct.method = "signature"
  

General purpose utilities

• cdxcore.verbose provides user-controllable context output for providing progress updates to users.

• cdxcore.util offers a number of utility functions such as standard formatting for dates, big numbers, lists, dictionaries etc.

• cdxcore.npio provides a low level binary i/o interface for numpy files.

• cdxcore.npshm provides shared memory numpy arrays.