run.py

import pandas as pd
import argparse

from src.preprocess import bound, negate, non_dimensionalize, deduplicate
from src.create     import build_coefficient_matrix
from src.eigs       import remove_inf_eigs
from src.io         import *

#
# filename: run.py
#
# @author: Alberto Serrano
#
# purpose:
#

DATA_FILENAME = "data/OB_crossslot_symmetric_De0.36.csv"
MTLB_FILENAME = "out/coefficient_matrices"

def init(args):
    """
    Initialize data by using preexisting dataframe in pickle file. If there is
    no file, read in the csv file and preprocess the data. Save the preprocessed
    data into a pickle file for later use.
    """

    # Extract the filename of the csv file from the path and remove the file
    # extension.
    filename = "".join(args["file"].split("/")[-1].split(".")[:-1])

    # Check if preprocessed data frame exists in intermediate_data directory
    if check_dataframe(filename) and args["use_pickle"]:

        print("Loading preexisting pickle file: {0}".format(filename))

        df = read_dataframe(filename)

    # If preprocessed data does not exist, load raw data and preprocess
    else:
        path = args["file"]
        print("Reading in file: {0}".format(path))

        # Create data frame with simulation data.
        # out.csv is pre-filtered data from a 90-degree bend flow.
        # Simulation properties include: De = 5, delta = 1e-2, shear banding
        df = pd.read_csv(path)

        # Bound data
        df = bound(df)

        # Fix values in data by negating those on right side
        negate(df)

        # Convert dimensional data into non-dimensionalized
        non_dimensionalize(df)

        # Remove duplicate values, typically occur on the line of symmetry
        # Updates indices
        deduplicate(df, up_index=True)

        # df["Boundary"] = [1] * df.shape[0]

        df["Points:0"] = df["Points:0"].round(3)
        df["Points:1"] = df["Points:1"].round(3)

        # Save dataframe for later use
        save_dataframe(df, filename)

    return df


def coefficient_matrix_setup(df):
    """
    Build the coefficient matrices and save for later use.
    """
    # Run with neighbor implementation
    A, B   = build_coefficient_matrix(df)

    # import matplotlib.pyplot as plt
    # import seaborn as sns
    #
    # sns.set()
    #
    # fig, axes = plt.subplots(figsize=(9,9), dpi=100)
    # axes.scatter(df["Points:0"], df["Points:1"], c=df["Boundary"], s=7)
    # axes.set_title("Cross slot with boundaries")
    # axes.set_xlabel("x")
    # axes.set_ylabel("y")
    #
    # plt.savefig("cross_bounds.png")
    # plt.show()

    # Remove infinite eigenvalues
    F, G_B = remove_inf_eigs(A, B)

    # Save coefficient matrices
    save_matrix_to_ml(F, G_B, MTLB_FILENAME)


def main():

    # Create argument parser
    ap = argparse.ArgumentParser(description="Create sparse matrix from data and more")

    # Add file argument
    ap.add_argument(
        "-f",
        "--file",
        type=str,
        help="Name csv file",
        default=DATA_FILENAME
    )

    ap.add_argument(
        "-v",
        "--verbose",
        help="increase output verbosity",
        action="store_true"
    )

    ap.add_argument(
        "-up",
        "--use-pickle",
        help="Ignore preexisting preprocessed data",
        action="store_true"
    )

    # Parse arguments
    args = vars(ap.parse_args())

    # Load data frame
    df = init(args)

    # Setup coefficient matrices and save.
    coefficient_matrix_setup(df)


main()