23

I am generating CSV files that needs to be opened and reviewed in Excel once they have been generated. It seems that Excel requires a different encoding than UTF-8.

Here is my config and generation code:

csv_config = {col_sep: ";", 
              row_sep: "\n", 
              encoding: Encoding::UTF_8
             }

csv_string = CSV.generate(csv_config) do |csv|
  csv << ["Text a", "Text b", "Text æ", "Text ø", "Text å"]
end

When opening this in Excel, the special characters are not being displayed properly:

Text a  Text b  Text æ Text ø Text å

Any idea how to ensure proper encoding?

Cjoerg
  • 1,271
  • 3
  • 21
  • 63
  • Try putting `# encoding: UTF-8` as your Ruby file's first line (second if you have a hash-bang line, `#!/usr/bin/env ruby`). I believe you are *writing* in UTF-8, but the Ruby source file is taken to be encoded as US_ASCII. (With Ruby 2.0+, source encoding defaults to UTF-8) – Amadan May 21 '15 at 08:30
  • I am using `ruby 2.0.0p247 (2013-06-27 revision 41674) [x86_64-darwin12.4.0]` so I suppose that means that my installation is already defaulting to UTF-8. – Cjoerg May 21 '15 at 09:40
  • 2
    No experience with Ruby. But Excel can open semicolon delimited CSV files which are UTF-8 encoded. But the file must have a BOM at its beginning. And whether the semicolon can be used as delimiter is locale dependent. So the best approach is to use tabulator delimited CSV and those UTF-16LE encoded. This should be most locale independent. – Axel Richter May 21 '15 at 12:13
  • What Excel are you using? I had no trouble getting the special characters to display in Excel 2013. – Fred Sep 10 '15 at 09:23
  • Another hint: With the [axlsx-gem](https://rubygems.org/gems/axlsx/) it is easy to create direct a xlsx-files. – knut Apr 20 '16 at 19:03

5 Answers5

32

Excel understands UTF-8 CSV if it has BOM. That can be done like:

Use CSV.generate

# the argument of CSV.generate is default string
csv_string = CSV.generate("\uFEFF") do |csv|
  csv << ["Text a", "Text b", "Text æ", "Text ø", "Text å"]
end

Use CSV.open

filename = "/tmp/example.csv"

# Default output encoding is UTF-8
CSV.open(filename, "w") do |csv|
  csv.to_io.write "\uFEFF" # use CSV#to_io to write BOM directly 
  csv << ["Text a", "Text b", "Text æ", "Text ø", "Text å"]
end
Stephen Rauch
  • 47,830
  • 31
  • 106
  • 135
cuzic
  • 498
  • 4
  • 6
31

The top voted answer from @joaofraga worked for me, but I found an alternative solution that also worked - no UTF-8 to ISO-8859-1 transcoding required.

From what I've read, Excel, can indeed handle UTF-8, but for some reason, it doesn't recognize it by default. But if you add a BOM to the beginning of the CSV data, this seems to cause Excel to realise that the file is UTF-8.

So, if you have a CSV like so:

csv_string = CSV.generate(csv_config) do |csv|
  csv << ["Text a", "Text b", "Text æ", "Text ø", "Text å"]
end

just add a BOM byte like so:

"\uFEFF" + csv_string

In my case, my controller is sending the CSV as a file, so this is what my controller looks like:

def show
  respond_to do |format|
    format.csv do
      #  add BOM to force Excel to realise this file is encoded in UTF-8, so it respects special characters
      send_data "\uFEFF" + csv_string, type: :csv, filename: "csv.csv"
    end
  end
end

I should note that UTF-8 itself does not require or recommend a BOM at all, but as I mentioned, adding it in this case seemed to nudge Excel into realising that the file was indeed UTF-8.

seymore_strongboy
  • 934
  • 10
  • 16
11

You should switch the encoding to ISO-8859-1 as following:

CSV.generate(encoding: 'ISO-8859-1') { |csv|  csv << ["Text á", "Text é", "Text æ"] }

For your context, you can do this:

config = {
  col_sep: ';',
  row_sep: ';',
  encoding: 'ISO-8859-1'
}

CSV.generate(config) { |csv|  csv << ["Text á", "Text é", "Text æ"] }

I had the same issue and that encoding fixed.

joaofraga
  • 608
  • 7
  • 17
  • The answer above worked for me, but only after I removed the `col_sep` and `row_sep` arguments. Just the `encoding: 'ISO-8859-1'` was all I needed. For context, the specific issue I was having was `é` characters appearing as `é` – seymore_strongboy Nov 02 '17 at 17:59
  • Good catch Greg, I will update the example without the context. – joaofraga Nov 02 '17 at 21:18
  • yeah, sure, 8859-1 good for everything yay, try writing some japanese or arabic characters in a CSV file like that. OP asked specifically for UTF-8, so why advise him to go decades back in time? – quetzalcoatl Jul 30 '20 at 22:29
0
config = {

  encoding: 'ISO-8859-1'
}

CSV.generate(config) { |csv|  csv << ["Text á", "Text é", "Text æ"] }
Sebastian Ruiz Perez
  • 93
  • 7
0

With https://github.com/gtd/csv_builder, I had to:

In the controller action:

@output_encoding = 'UTF-8'
send_data "\uFEFF" + render_to_string(), type: :csv, filename: @filename

Atop the csv.csvbuilder template:

faster_csv.to_io.write("\uFEFF")

I don't know why I had to add the BOM twice, but it did not work with either one on its own.

Steve Friedman
  • 21
  • 2