Arcpy vs Geopandas

flow.py

import sys

import geopandas

from helpers import width, peak_flow


def flow_column_name(return_period):
    return "Q{:d}".format(int(float(return_period) * 10))


def populate_column(table, column, fxn, **kwargs):
    assignment = {column: lambda df: df.apply(fxn, axis=1, **kwargs)}
    return table.assign(**assignment)


def compute_width_and_flow(filename):
    # list of all storm event return periods
    return_periods = ['50', '20', '10', '4', '2', '1', '0.5', '0.2']
    flow_columns = list(map(flow_column_name, return_periods))

    width_col = 'width'
    result_columns = flow_columns + [width_col]

    # load the attribute table
    table = geopandas.read_file(filename)

    # compute the channel width
    table = populate_column(table, "width", width)

    # compute the flows
    for fc in flow_columns:
        table = populate_column(table, fc, peak_flow, colname=fc)

    # write to file
    table.to_file(filename)

    return table


if __name__ == '__main__':
    compute_width_and_flow(sys.argv[1])

flow_arc.py

import sys

import numpy
import arcpy

from helpers import width, peak_flow


def compute_width_and_flow(filename):
    # list of all storm event return periods
    return_periods = ['50', '20', '10', '4', '2', '1', '0.5', '0.2']
    flow_columns = list(map(flow_column_name, return_periods))

    width_col = 'width'
    result_columns = flow_columns + [width_col]

    # load the attribute table
    table = load_attr_table(filename, cols_to_exclude=result_columns)

    for col in result_columns:
        table = add_column_to_table(table, column_name, -9999.9)

    update_attr_table(filename, table)
    populate_column(filename, "width", width, keyfields=['DRNAREA', 'AggregateEcoregion'])

    for rp in return_periods:
        column_name = flow_column_name(rp)
        other_fields = ['AggregateEcoregion', column_name, 'DRNAREA', 'PRECIP', 'ELEV']
        populate_column(filename, rp, peak_flow)
    return table


def load_attr_table(filename, cols_to_exclude):
    fields = filter(
        lambda x: x.type != 'Date' and x.name not in cols_to_exclude,
        arcpy.ListFields(filename)
    )
    fieldnames = [f.name for f in fields]
    return arcpy.da.FeatureClassToNumPyArray(in_table=filename, field_names=fieldnames)


def update_attr_table(filename, table, id_col, orig_columns, new_columns):
    all_columns = [id_col] + orig_columns

    with arcpy.da.UpdateCursor(filename, all_columns) as cur:
        for oldrow in cur:
            newrow = find_row_in_table(table, id_col, oldrow[0])
            if newrow is not None:
                for n, col in enmumerate(new_columns, 1):
                    oldrow[n] = newrow[col]

            cur.updateRow(oldrow)

    return filename


def find_row_in_table(table, column, value):
    rows = filter(lambda x: x[column] == value, table)
    if len(rows) == 0:
        row = None
    elif len(rows) == 1:
        row = row[0]
    else:
        raise ValueError('too many rows where {} == {}'.format(column, value))
    return row


def flow_column_name(return_period):
    return "Q{:d}".format(int(float(return_period) * 10))


def populate_column(table, column, fxn, **kwargs):
    newtable = table.copy()
    for ii, row in enumerate(table):
        newtable[ii][column] = fxn(row, **kwargs)

    return newtable


def add_column_to_table(array, new_column, new_values):
    from numpy.lib.recfunctions import append_fields

    # convert the new value to an array if necessary
    if numpy.isscalar(new_values):
        new_values = numpy.array([new_values] * array.shape[0])

    # append the new colum
    new_array = append_fields(array, new_column, new_values)

    return new_array.data


if __name__ == '__main__':
    compute_width_and_flow(sys.argv[1])

This is interesting. I don't use ArcPy but would be curious to see what it's performance is like and if it leverages vectorization to process spatial data more quickly than GeoPandas.