One-line queries of the SuiteSparse matrix collection 21 Jun 2022

ssquery.py

#!/usr/bin/env python3
""" Usage:
python3 ssquery.py  '(posdef == 1)  &  (1000 <= rows <= 20000)'  posdef.csv
reads http://sparse.tamu.edu/files/ssstats.csv 
queries it
writes "posdef.csv" like

# SuiteSparse  posdef  &  (1000 <= rows <= 20000) 
id,group,name,rows,cols,nnz,real,bool,is2d3d,posdef,symm,nsymm,kind
29156,Boeing,msc01050,1050,1050,26198,1,0,1,1,1.0,1.0,structural
...
554466,JGD_Trefethen,Trefethen_20000,20000,20000,554466,1,0,0,1,1.0,1.0,combinatorial

Then look at the query csv with your favorite tools
(awk grep sort editor spreadsheet),
download matrices with e.g. https://github.com/drdarshan/ssgetpy

More examples:
    '(symm == 1)  &  (posdef == 0)'  tmp.csv (the default)
    '(symm == 1)  &  (nnz / rows <= 4)'

Note that id s are not unique: 48 50 76 149 238 306 ...
"""

# Keywords: sparse-matrix SuiteSparse MatrixMarket python pandas pandas-query

# why ? flexible queries with python pandas
# wibni: a column "year"
# wibni: a csv / wiki of eigval info -- emin, emax, enear0

import os
import pandas as pd  # https://pandas.pydata.org/docs/user_guide

#...............................................................................
url = "http://sparse.tamu.edu/files/ssstats.csv"
_ssinfo = None  # load_ssinfo -> this pandas DataFrame -> ssquery


def ssquery( q = "(posdef == 1)  &  (1000 <= rows <= 20000)" ) -> pd.DataFrame:
    """ https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.query.html """
    global _ssinfo  # class ? mttiw
    if _ssinfo is None:
        _ssinfo = load_ssinfo()
    return _ssinfo.query( q )


def load_ssinfo( csvin=url, **keywords ) -> pd.DataFrame:
    """ read_csv %s  or a local .csv """ % url
        # square only, drop duplicates
    csvin = _url_localfile( csvin )
    df = pd.read_csv( csvin,
        names="group name rows cols nnz real bool is2d3d posdef symm nsymm kind id ".split(),
        index_col="id",
        skiprows=2,  # 2893  08-Oct-2020 17:09:58
        sep=",",  # whitespace: "\\s+"
        comment="#",
        **keywords
        )
    df = df[ df.rows == df.cols ]  # square only
    df = df[ ~ df.kind.str.contains( "duplicate" )]  # drop
    df.kind = (df.kind.str.replace( " problem", "" )
                .str.replace( "computational fluid dynamics", "cfd" )
                .str.replace( " ", "-" ))
    df = df.sort_values( "rows", kind="stable" )
    # print( "read %s  %s \n" % ( csvin, df ))
    return df


def write_csv( csvout: str, df: pd.DataFrame, header="", **keywords ):
    """ df.to_csv with #header """
    nr, nc = df.shape
    print( "\nwrite_csv %s: %d rows x %d columns" % (
            csvout, nr, nc ))
    with open( csvout, "w" ) as f:
        if header:
            f.write( "# %s \n" % header )  # good csvs should have headers
                # + date pwd: pandasutil.py
        df.to_csv( f, **keywords )


def _url_localfile( url ) -> "tail if isfile  else url":
    if os.path.isfile( url ):
        return url
    tail = os.path.split( url )[1]  # after last /
        # findfile( tail, dirs=". data $data ~/.data" )
    return tail if os.path.isfile( tail ) \
        else url

__version__ = "2022-06-21 June  denis-bz-py t-online.de"

#...............................................................................
if __name__ == "__main__":  # python ssquery.py 'query'  runs the following --
    import sys

        # default args --
    query = "(posdef == 1)  &  (1000 <= rows <= 20000)"
        # "posdef" alone ??
    csvout = "tmp.csv"

    args = sys.argv[1:]
    if args:
        if args[0].startswith( ("-h", "--h") ):
            print( __doc__ )
            sys.exit()
        query = args.pop( 0 )
        if args:
            csvout = args.pop( 0 )

    #...........................................................................
    querydf = ssquery( query )
    print( "\n-- query '%s' \n%s \n" % (
            query, querydf ))
    if csvout:
        write_csv( csvout, querydf, header="SuiteSparse  " + query )
    # test = pd.read_csv( csvout, index_col="id", sep=",", comment="#" )