Skip to content

Latest commit

 

History

History
93 lines (80 loc) · 5.36 KB

README_EN.md

File metadata and controls

93 lines (80 loc) · 5.36 KB
Raw

bkflow-dmn: A DMN (Decision Model Notation) library for Python

Introduction

bkflow-dmn is a Python-based DMN (Decision Model Notation) library that uses FEEL (Friendly Enough Expression Language) as its description language. It can serve as a decision engine to solve decision problems in real business scenarios.

Quick Start

1. Install dependencies

$ pip install bkflow-dmn

2. Construct decision table data

Currently, bkflow-dmn supports single table decisions. It allows describing decision tables using Python dictionaries, as shown in the example below:

decision_table = {
    "title": "decision table", # Table title, currently has no practical use
    "hit_policy": "Unique", # Hit policy
    "inputs": { # Input table
        "cols": [{"id": "score"}, {"id": "sex"}], # Column descriptions
        "rows": [ # Values corresponding to rows
            ["[0..70]", '"boy"'],
            ["(70..80]", '"boy"'],
            ["(80..90]", '"boy"'],  # For actual scenarios, if you want to [represent multiple column expressions in a single row], use a string type description, like 'score in (80..90] and sex="boy"'
            ["(90..100]", '"boy"'],
            ["[0..60]", '"girl"'],
            ["(60..70]", '"girl"'],
            ["(70..80]", '"girl"'],
            ["(80..100]", '"girl"'],
        ],
    },
    "outputs": { # Output table
        "cols": [{"id": "level"}], # Column descriptions
        "rows": [ # Values corresponding to rows
            ['"bad"'],
            ['"good"'],
            ['"good+"'],
            ['"good++"'],
            ['"bad"'],
            ['"good"'],
            ['"good+"'],
            ['"good++"'],
        ],
    },
}

The input table defaults to evaluating the expression in each cell as a whole for truth value determination. It also supports some shorthand notations (only when not using [single row to represent multiple column expressions]):

Shorthand Completed expression (assuming column name is x)
">=10" "x>=10"
">10" "x>10"
"<=10" "x<=10"
"<10" "x<10"
"[1..2]" "x in [1..2]"
"(1..2]" "x in (1..2]"
'"boy"' 'x="boy"'
"1" "x=1"
"1.2" "x=1.2"
"-1" "x=-1"
"-2.3" "x=-2.3"
"true" "x=true"
"false" "x=false"
"null" "x=null"

3. Call the function to make decisions

from bkflow_dmn.api import decide_single_table

facts = {"score": 10, "sex": "boy"} # Column data and context for the input table
result = decide_single_table(decision_table, facts, strict_mode=True)  # [{"level": "bad"}]

For multi-table decisions, you can currently combine it with the process engine bkflow-engine.

4. Supported hit policies

Hit Policy Description
Unique Unique hit, in strict mode, only one rule should be hit
Any Any hit, in strict mode, there must be a hit result and the hit result must be unique
First First hit, in strict mode, there must be a rule hit
Priority Priority hit, using output as sorting priority, in strict mode, there must be a rule hit and the priority sorting result must be unique
RuleOrder Multiple rule hit policy, output according to the order of rule definition
OutputOrder Multiple output hit policy, sort and output according to the output order
Collect Multiple output hit policy, output all matching results
Collect(Sum) Multiple output hit policy, sum the output values
Collect(Max) Multiple output hit policy, find the maximum of the output values
Collect(Min) Multiple output hit policy, find the minimum of the output values
Collect(Count) Multiple output hit policy, count the number of output values