rh-marketingops / dwm Goto Github PK

UDFs should be defined and executed as extensions of a base UDF class:

from dwm import UDF

class MyUDF(UDF):
  udf_name = 'publicly-visible name' 
  udf_docs = 'publicly-visible docs'
  def run(self, data, hist):
  ...

Currently, the DWM config dictionary is passed down to each sub-function, as is the MongoDB config.
We need to refactor this to have a top-level DWM object which is configured once, and all subsequent calls reference.

Example:

from dwm import DwmInstance

DWM = DwmInstance()
DWM.set_mongo('mongo_connection_string')
DWM.set_config({...}) # pass in dictionary with config object

INPUT = {...} # input record for cleaning

OUTPUT, HISTORY = DWM.clean(INPUT)

We should be able to preserve most, if not all, of the low-level functions used in DWM already; this would just replace how we interact with it.

Option to skip deriving value on a field if existing value in picklist of acceptable values.

Currently loop through config.derive settings breaks only when the actual input field value is changed, which leads to inconsistent behavior. Need to instead return a boolean flag from the child functions which would break the loop.

Currently, running dwmOne results in the contactHistory record being auto-added to MongoDB. To make this package more scalable as a microservice, we need to return the history record separately.

Example:

INPUT = {...} # single record to clean
OUTPUT, HISTORY = dwmOne(INPUT, ...)

This update should include removing _current, timestamp, and the key identifier from history record.

Add option for derive logic to skip if the value is already valid and/or clear any non-valid values.

When records are added to contactHistory, include custom metadata for better database cleanup/management.

{
    "IdentifierKey": "12345",
    "metadata": {
        "batchId": "abcde",
        "batchTimestamp": timestamp
    },
    ...

Following the major refactor for v0.1.0, make the low-level functions match the patterns established in the high-level wrappers

In Readme.md there is a small typo. Should it be Usage instead of Useage?