Pandas Block Model to OMF

This example demonstrates how to read a block model from an OMF file, modify it using pandas, and write it back to a new OMF file. A round trip is completed to validate that the omf file stores various data types as expected.

from pathlib import Path

import numpy as np
import pandas as pd
import omf
from omf import Project

from omfpandas import OMFPandasReader, OMFPandasWriter
from omfpandas.blockmodel import df_to_blockmodel

Instantiate

Read the original block model from the OMF file.

test_omf_filepath: Path = Path('../assets/v2/test_file.omf')
blocks: pd.DataFrame = OMFPandasReader(filepath=test_omf_filepath).read_blockmodel(blockmodel_name='vol')
blocks.head()

						random attr
x	y	z	dx	dy	dz
10.5	10.5	-9.5	1.0	1.0	1.0	0.727986
11.5	10.5	-9.5	1.0	1.0	1.0	0.277389
12.5	10.5	-9.5	1.0	1.0	1.0	0.351741
13.5	10.5	-9.5	1.0	1.0	1.0	0.999272
14.5	10.5	-9.5	1.0	1.0	1.0	0.495092

Modify the BlockModel

We’ll modify the block model using Pandas and write it back to another OMF file. Add a new float column to the block model.

blocks['density'] = 2.7
# set the second density record to nan
blocks.iloc[1, blocks.columns.get_loc('density')] = np.nan

Create an integer variable

np.random.seed(42)
blocks['some_int'] = pd.Series(np.random.choice([0, 1, 2, np.nan], size=blocks.shape[0], ), dtype='Int32',
                               index=blocks.index)

Create a categorical variable

blocks['rock_class'] = np.random.choice(['A', 'B', 'C', np.nan], size=blocks.shape[0])
blocks['rock_class'] = blocks['rock_class'].astype('category')
blocks.head()

						random attr	density	some_int	rock_class
x	y	z	dx	dy	dz
10.5	10.5	-9.5	1.0	1.0	1.0	0.727986	2.7	2	nan
11.5	10.5	-9.5	1.0	1.0	1.0	0.277389	NaN	<NA>	B
12.5	10.5	-9.5	1.0	1.0	1.0	0.351741	2.7	0	B
13.5	10.5	-9.5	1.0	1.0	1.0	0.999272	2.7	2	nan
14.5	10.5	-9.5	1.0	1.0	1.0	0.495092	2.7	2	A

blocks.dtypes

random attr     float64
density         float64
some_int          Int32
rock_class     category
dtype: object

Write to OMF

new_omf_filepath: Path = Path('modified_test_file.omf')

bm = df_to_blockmodel(blocks, 'Modified Block Model')

OMFPandasWriter(filepath=new_omf_filepath).write_blockmodel(blocks=blocks, blockmodel_name='Modified Block Model',
                                                            allow_overwrite=True)
omfpr: OMFPandasReader = OMFPandasReader(filepath=new_omf_filepath)

2024-11-18 12:21:02,533 - INFO - Writing dataframe to BlockModel: Modified Block Model
2024-11-18 12:21:02,547 - INFO - Elapsed time for write_blockmodel: 0:00:00.013654

Confirm that the new variables are in the model we saved.

saved_blocks: pd.DataFrame = OMFPandasReader(filepath=new_omf_filepath).read_blockmodel(
    blockmodel_name='Modified Block Model')
saved_blocks.head()

						random attr	density	some_int	rock_class
x	y	z	dx	dy	dz
10.5	10.5	-9.5	1.0	1.0	1.0	0.727986	2.7	2	nan
11.5	10.5	-9.5	1.0	1.0	1.0	0.277389	NaN	<NA>	B
12.5	10.5	-9.5	1.0	1.0	1.0	0.351741	2.7	0	B
13.5	10.5	-9.5	1.0	1.0	1.0	0.999272	2.7	2	nan
14.5	10.5	-9.5	1.0	1.0	1.0	0.495092	2.7	2	A

saved_blocks.dtypes

random attr     float64
density         float64
some_int         object
rock_class     category
dtype: object

Validation

Load the two omf projects and compare the attribute data for each

project_1: Project = omf.load(str(test_omf_filepath))
project_2: Project = omf.load(str(new_omf_filepath))

bm_1: omf.TensorGridBlockModel = [element for element in project_1.elements if element.name == 'vol'][0]
bm_2: omf.TensorGridBlockModel = [element for element in project_2.elements if
                                  element.name == 'Modified Block Model'][0]

for i, attr_1 in enumerate(bm_1.attributes):
    attr_2 = bm_2.attributes[i]
    assert np.all(attr_1.name == attr_2.name)
    assert np.all(attr_1.array.array == attr_2.array.array)