JMP18: python: conversion from pandas to dataTable skips columns with missing values

[#1]

import jmp, numpy as np, pandas as pd
import jmputils

def p2j(df):
	dt2 = jmp.DataTable('dt2', df.shape[0])
	# get the column names from the data frame
	names = list(df.columns)
	# loop across the columns of the data frame
	for j in range( df.shape[1]):
		# check if the coulumn data type is string or numeric
		if is_string_dtype(df[names[j] ]):
			dt2.new_column(names[j], jmp.DataType.Character)
		else:
			dt2.new_column(names[j], jmp.DataType.Numeric)
		# populate the JMP column with data
		dt2[j] = list(df.iloc[:,j])
	return(dt2)

fname = r"export.csv";
df = pd.read_csv(fname)
filt_csv = p2j(df);

Using above code to read in a csv file. The p2j function seems to not import col3 correctly as it has missing values

csv file contents

"col1","col2","col3"
"a","b","c"
"a","b",""
"a","b","e"
"a","b",""

[#2]

The short answer is that is_string_dtype( col3 ) returns False.  It is not a string column, it is a column of objects, some are strings some are NaN.  The code above treats everything that is not a string as numeric. The NaN's go in properly as missing, but the character values 'c' and 'e' are invalid for a numeric column and become missing values.  I'm looking for a better answer on reading the pandas column as character, without testing every value.

[#3 · Accepted solution]

Updated code shown here, if the column type is not numeric or string, check until we get a non-missing value before determining the column type as character or numeric.

import jmp
import numpy as np, pandas as pd
from pandas.api.types import is_string_dtype
from pandas.api.types import is_numeric_dtype

def p2j(df):
	dt2 = jmp.DataTable('dt2', df.shape[0])
	# get the column names from the data frame
	names = list(df.columns)
	# loop across the columns of the data frame
	for j in range( df.shape[1]):
		# check if the coulumn data type is string or numeric
		if is_numeric_dtype( df[ names[j] ] ):
			dt2.new_column(names[j], jmp.DataType.Numeric)
			dt2[j] = list(df.iloc[:,j])			# populate the JMP column with data
		elif is_string_dtype( df[ names[j]] ):
			dt2.new_column(names[j], jmp.DataType.Character)
			dt2[j] = list(df.iloc[:,j])			# populate the JMP column with data
		else:
			col = list()
			dtype = None
			for i in range(dt2.nrows):
				cell_value = df.iloc[i,j]
				# delay setting column type until we see first non-null value
				if pd.isna( cell_value ):
					col.append(None)
				else:
					if not dtype:
						if is_numeric_dtype( cell_value ):
							dtype = jmp.DataType.Numeric
						else:
							# assume it's character - could be datetime64,... 
							dtype = jmp.DataType.Character
					col.append( cell_value )
			print(col)
			if dtype: 
				dt2.new_column(names[j], dtype)
				dt2[j] = col				# populate the JMP column with data
		
	return(dt2)

fname = r"export.csv";
df = pd.read_csv(jmp.HOME + 'Downloads/' + fname)
print(df)
print(df.dtypes)
print( f'col1 is a string: {is_string_dtype(df[names[0]])}' )
print( f'col2 is a string: {is_string_dtype(df[names[1]])}' )
print( f'col3 is a string: {is_string_dtype(df[names[2]])}' )

filt_csv = p2j(df);

This gives the results:

/*:

  col1 col2 col3
0    a    b    c
1    a    b  NaN
2    a    b    e
3    a    b  NaN


col1    object
col2    object
col3    object
dtype: object


col1 is a string: True


col2 is a string: True


col3 is a string: False


['c', None, 'e', None]

Notice in my function I converted the NaN which is numeric to None.  The JMP DataTable code will accept None either in Numeric columns or Character columns and does the 'right' thing.  By putting a NaN in a numeric column and "" empty string in a Character column.

[#4]

Using the above p2j( ) and an updated export.csv with the a column of numeric data having a missing value as the first value gives:

% more ~/Downloads/export.csv

"col1","col2","col3","col4"

"a","b","c",

"a","b","",2.718

"a","b","e",1.618

"a","b","",3.14159