polarmints 0.1.24

syntactic sugar and additional namespaces for polars

Polarmints

Syntactic sugar for polars
Apologies, not all features documented so feel free to explore codebase

Extensions

extends polars Dataframes with additional namespaces for convenience functions
example:

import polars as pl
from polarmints import PolarMints, c, DF
__all__ = [PolarMints] # required for extending DFs with polarmints, even though not explicitly used  

df = DF({
    'a': [1, 2, 3],
    'b': [1, 2, 3],
})

df2 = DF({
    'a': [1, 2, 3],
    'c': [1, 2, 3],
}, schema_overrides={'a': pl.Int16})

# df.pm: convenience helper funcs
joined = df2.pm.join(df, 'a') # implicitly converts datatypes before joining two DFs whose column types don't match

# this is contrived example since it's more efficient to do in polars: pl.DataFrame.with_column(pl.col('a') + 1) 
# however pandas may have other dataframe and series methods not yet implemented in polars
added_col = df.pd.assign(a2=1)

DAG

Given an input pl.DataFrame each @node decorated method on a SubClass of DagBase represents a derived column which could themselves depend on other derived columns. A dag is required to represent this hierarchy of dependencies, i.e. which columns to derive first and which ones can be done in parallel. this framework is inspired by MDF and the gromit dag in beacon.io except the nodes represent polars expressions instead of plain python.

Example usage :

from polarmints.dag.core import DagBase, node, s
from polarmints import c, DF

class DagExample(DagBase):

    @node
    def DerivedCol(self):
        return c['raw2'] + 2

    @node
    def OverridenCol(self):
        """
        input column with this name will be overridden by this method if instance is initialized with
        override_existing=True
        """
        return c['raw1'] + 1

    @node
    def DerivedCol_2ndOrder(self):
        """
        NOTE: 's' and 'c' are effectively the same, 's' is merely for readability to distinguish derived columns (s)
        from raw inputs (c)
        """
        return s['OverridenCol'] + c['raw3']

    @node
    def DerivedCol_2ndOrder_B(self):
        return s['OverridenCol'] + s['DerivedCol']


if __name__ == '__main__':
    # this is an instance instead of class because some usages may require initializing the dag with instance specific
    # params when multiple instances are used in the same process.
    example = DagExample()

    # mock inputs
    df = DF({
        'raw1': [1, 2, 3],
        'raw2': [1, 2, 3],
        'raw3': [1, 2, 3],
        'OverridenCol': [10, 11, 12]
    })

    # select desired derived columns from mock inputs using dag
    df1 = example.with_cols(df,
        # func siganture: *args and **kwargs expresisons behave the same way as pl.DataFrame.with_column() and .select()          
        example.DerivedCol_2ndOrder,
        example.OverridenCol, #this will not be overridden
        'raw2',  # can be mixed with raw pl.Exprs that don't depend on the DAG nodes
        c['raw3'] + 2,
        
        **{
            'd1': example.DerivedCol,
            'd2': example.DerivedCol_2ndOrder_B,
            'd3': c['raw1'] * c['raw2']
        },
    )
    print(df1)

    """
    shape: (3, 8)
    ┌──────┬──────┬──────┬──────────────┬─────────────────────┬─────┬─────┬─────┐
    │ raw1 ┆ raw2 ┆ raw3 ┆ OverridenCol ┆ DerivedCol_2ndOrder ┆ d1  ┆ d2  ┆ d3  │
    │ ---  ┆ ---  ┆ ---  ┆ ---          ┆ ---                 ┆ --- ┆ --- ┆ --- │
    │ i64  ┆ i64  ┆ i64  ┆ i64          ┆ i64                 ┆ i64 ┆ i64 ┆ i64 │
    ╞══════╪══════╪══════╪══════════════╪═════════════════════╪═════╪═════╪═════╡
    │ 1    ┆ 1    ┆ 1    ┆ 10           ┆ 11                  ┆ 3   ┆ 13  ┆ 1   │
    │ 2    ┆ 2    ┆ 2    ┆ 11           ┆ 13                  ┆ 4   ┆ 15  ┆ 4   │
    │ 3    ┆ 3    ┆ 3    ┆ 12           ┆ 15                  ┆ 5   ┆ 17  ┆ 9   │
    └──────┴──────┴──────┴──────────────┴─────────────────────┴─────┴─────┴─────┘
    """

    # another example with more params yielding more implicitly derived columns
    expressions = [
        example.DerivedCol_2ndOrder, example.DerivedCol_2ndOrder_B,
    ]
    df2 = example.select(df, 'raw2', *expressions,
         include_deps=True, # include intermediate dependencies as columns in result DF for higher order nodes
         override_existing=True, # override the existing column if dict key or node name conflicts with raw input column
    )
    print(df2)

    """
    shape: (3, 5)
    ┌──────┬────────────┬──────────────┬───────────────────────┬─────────────────────┐
    │ raw2 ┆ DerivedCol ┆ OverridenCol ┆ DerivedCol_2ndOrder_B ┆ DerivedCol_2ndOrder │
    │ ---  ┆ ---        ┆ ---          ┆ ---                   ┆ ---                 │
    │ i64  ┆ i64        ┆ i64          ┆ i64                   ┆ i64                 │
    ╞══════╪════════════╪══════════════╪═══════════════════════╪═════════════════════╡
    │ 1    ┆ 3          ┆ 2            ┆ 5                     ┆ 3                   │
    │ 2    ┆ 4          ┆ 3            ┆ 7                     ┆ 5                   │
    │ 3    ┆ 5          ┆ 4            ┆ 9                     ┆ 7                   │
    └──────┴────────────┴──────────────┴───────────────────────┴─────────────────────┘
    """

    # for debugging: examine which derived expressions can be evaluated in parallel for each step
    ordered_exprs = example.ordered_exprs(expressions)
    print([[str(e) for e in oe] for oe in ordered_exprs])

    """
    [
        [
            '[(col("raw1")) + (1)].alias("OverridenCol")', 
            '[(col("raw2")) + (2)].alias("DerivedCol")'
        ], [
            '[(col("OverridenCol")) + (col("raw3"))].alias("DerivedCol_2ndOrder")',
            '[(col("OverridenCol")) + (col("DerivedCol"))].alias("DerivedCol_2ndOrder_B")'
        ]
    ]
    """