fancy way to load contents of a CSV file into a dictionary<string,string> in C#

Question

I know how to do this via streamreader and read to the end of line, but was curious if there was a fancier way of going it (for the sake of learing).

filename: blah.csv

File layout is simple:

"Some234234 ", 234
"blahblha234234 ", 2322

I want to load this into a dictionary (the second part should be a int, but I will parse later in case of errors).

Can you do this via linq somehow?

Answer 1

You should use the TextFieldParser class in Microsoft.VisualBasic.dll. (There's nothing wrong with using it in C#.

Remember that CSV is a much more complicated file format than you think; both quotes and newlines can be escaped.

• richard: Very true, and I have used it many times this way for complex csv's or csv's with unpredictable contents. But this is only necessary if you expect a complex file. In some cases it's a very simple file, for example all integer values.

Answer 2

CSV can be a lot more complicated than it appears at first. Here is an excellent library for reading CSV files.

http://www.codeproject.com/KB/database/CsvReader.aspx

• Aaronaught: I agree with this and with SLaks's reply. The most effective way to solve any problem is to use a solution that already exists. CSV seems to be one of those features that's just simple enough for people to re-invent on a whim, but still complex enough for most of them to make a mess out of.

Answer 3

Untested, but:

static IEnumerable<string> ReadLines(string path) {
 using (var file = File.OpenText(path)) {
 string line;
 while ((line = file.ReadLine()) != null) {
 yield return line;
 }
 }
}
var data = (from line in ReadLines(path)
 select line.Split(','))
 .ToDictionary(
 arr => arr[0].Trim('"', ' '),
 arr => int.Parse(arr[1].Trim()));

• SLaks: This will not handle escaped CSV files. Even if the CSV files aren't escaped right now, they might be in the future.

Answer 4

foreach (string s in File.ReadAllLines(filename)) {
 var vals = s.Split(',')
 dictionary[vals[0]] = vals[1];
}

No LINQ but this is simple really. Doesn't handle embedded ',''s in the first value though.

• SLaks: This will not handle escaped CSV files. Even if the CSV files aren't escaped right now, they might be in the future.
• Ron Warholic: He obviously doesn't care about things like this as he put it "for the sake of learing." CSV isn't even a standardized format, his CSV might not allow escaping.

Answer 5

 string[] file = File.ReadAllLines(@"C:\temp\dictionary.txt");
 Dictionary<string, string> b = (from p in file
 let x = p.Split(',')
 select x).ToDictionary(a => a[0], a => a[1]);

• SLaks: This will not handle escaped CSV files. Even if the CSV files aren't escaped right now, they might be in the future.

Answer 6

You could use File.ReadAllLines(), Select() and ToDictionary() to do this:

var d = File.ReadAllLines(file).Select( l => {
 var split = l.Split(',');
 return new { Key = split[0], Value = split[1] };
} ).ToDictionary( p => p.Key, p => p.Value );

But there are obvious problems here with respect to error handling and robustness, and as you add those, it gets worse and worse. I don't feel there's a particular good reason prefer LINQ or its extension methods here, as it isn't buying you much. The straightforward ways, posted already, are much cleaner.

EDIT: Sid's answer, for example, contains essentially the same code, but written in a much, much cleaner form by avoiding all this "fancy" junk.

Answer 7

This might be overkill in your simple scenario, but for CSV to strongly typed collection conversion, I usually use FileHelpers.

It's a great tool to have in the tool box.

• Piotr Kula: How can i use filehelpers to load CSV into the pregenereterd OR classes that can be used in LINQ2SQL instead of creating new class and assigning each damn field to it? I have a old school table with 300+ columns.. yay
• Martin R-L: I don't know your "pregenerated OR classes", but if minimizing code and violate SRP is okay, I guess you could use the same types and add the appropriate FileHelpers attributes. Personally, I'd separate the concerns, and map from the FileHelper types to the L2S types with AutoMapper.