[Tcllib-devel] Trying to deal with tcl and csv files

[Larry W. V. <lv...@gm...>]

So, I have a file that generates text of the form:

itemname;relationship name;date;user;"item 2"
itemname2;relationship name;date;user;"item 3
item 4
item 5"


and so forth.

I am wanting to write some code that reads these csv records in, one at a
time, and then loop through the list in the last column creating a series
of lines of the format

itemname --> list item 0;
itemname --> list item 1;
:
itemname -> list item n;

I tried to extract just column 1 and 4 from this file using the example
program csvcut - however, it did not handle the multiple line properly.

I read over the docs, and didn't see an input function - just the
iscomplete function.

so i wrote this:

package require Tcl
package require csv

proc readcsvrecord { fd } {
    gets $fd buffer
    # puts stderr "rcr 1: $buffer"
    while { [csv::iscomplete $buffer ]  == 0}  {
         gets $fd buff2
         set buffer ${buffer}${buff2}
         # puts stderr "rcr 2: $buffer"
        }
        return $buffer
}

puts "graph ci {"
set next [readcsvrecord stdin]
# puts stderr "eof returns [eof stdin]"

while { ! [eof stdin] } {
        # puts stderr "Record is $next"

        set l [csv::split $next "\;"]
        # puts stderr "list is $l"

        # error here
        foreach item [lindex $l 4] {
                puts "[lindex $l 0]->$item;"
        }
        set next [readcsvrecord stdin]
}
puts "}"
~


This sort of works. However, when generating my output (see "error here") I
have a problem when the item has white space in it - the white space is
turning one item into multiples.

Two questions:
1. Am I missing anything in the handling of the actual csv file?
2. Am I forgetting something basic to Tcl in handling the csv column
containing white space? I presume this is nothing csv specific, but instead
something basic that I have forgotten since I don't write a lot of code in
Tcl any more... sigh.

I really want to make certain my handling of the CSV records, and columns,
is correct. I was concerned about whether the backslash was needed before
the separator character; I didnt want Tcl to choke when it sees it.