Re: CPYFRMIMPF SQ99999 code 4

I honestly do not mean to spam the lists with Python recommendations,
but this is really a perfect place to use Python. It has a built-in
CSV module so the parsing is robust and dead-simple; and if you use
iSeriesPython, you can write directly to a PF with either RLA or SQL.

Using the stream file APIs from RPG is certainly fine, but... well, it
involves more boilerplate and more fumbling around if you haven't
already built a CSV-parsing service program.

Using iSeriesPython for this is stupidly easy. Unfairly easy. Here, I
will give you the code:

### begin script

import csv
from operator import itemgetter

import db2 # this is specific to iSeriesPython; for other Pythons,
use ibm_db_dbi instead

# Pick the 14 columns by ordinal number, using 0 for the first column.
selected_columns = itemgetter(0, 1, 2, 5, 7, 9, 11, 12, 51, 104, 105,
108, 120, 134)
placeholders = ', '.join('?' * 14)

connection = db2.connect(autocommit=True)
c1 = connection.cursor()

# Get the data types of the columns.
c1 = execute('select * from mylib.mytable')
types = [str(t[1]) for t in c1.description]

with open('mydata.csv') as f:
reader = csv.reader(f)
next(reader) # consume header; if no header, remove this line
for row in reader:
elements = []
for val, t in zip(selected_columns(row), types):
elements.append(val if t == 'STRING' else float(val))
c1.execute('''
insert into mylib.mytable values ({})
'''.format(placeholders), elements)

### end script

If you know ahead of time that all the columns you want are string, or
all numeric, then you can simplify the code.

If you use a Python other than iSeriesPython, then there are other
(relatively minor) adjustments that need to be made.

I'm happy to help with explanations or code adjustments.

John Y.