Experimental project to find the best way to ensure a preferred encoding in Strings coming from untrusted sources.
The most sane way of dealing with character data is choosing an internal representation for your application and convert all incoming and outgoing data when necessary.
To ensure that our internal encoding is always at least valid we can choose to do one of two things.
-
Throw an exception when we receive invalid character data. (ie. Our internal encoding is UTF-8 and we receive Latin-1 data or invalid UTF-8)
-
Accept whatever we get and try to mold it in such a way that it becomes usable in our application.
It is generally accepted to go for the second option, most of the times the end-user has no way of solving these problems because a vendor made a mistake. It’s not very nice to shut them out.
There are a number of techniques when molding the character data to our needs. Sniffing encoding, transcoding, dropping invalid characters are just a few examples. We implement a number of these techniques so you can easily protect your application from bad data.
We’ve crammed at lot of functionality into the ensure_encoding method because we want to keep the number of new methods on String to a minimum. We’ll walk through an example to show how it works.
example = 'Café' example.encoding => #<Encoding:ISO-8859-1>
After ensuring the encoding of a string you can at least assume that you can concatenate the string to another string with the same encoding. In other words, it contains data valid for the specified encoding.
example.ensure_encoding('UTF-8')
example.encoding => #<Encoding:UTF-8>
Beyond this you can specify a number of options to perform more operations to make sure the data in the string didn’t become unreadable garbage. Let’s look at a number of situations.
For instance, when you’re excepting data from browsers you can be pretty sure the character data is properly encoded. When someone does send bad data it’s probably a hacker and you can discard the request.
example.ensure_encoding('UTF-8',
:external_encoding => 'UTF-8,
:invalid_characters => :raise
)
In this scenario you’re connecting to web API through a ReST library and you know the encoding of the source data because it’s in the headers. However you’re not sure the encoding of the strings is valid.
example.ensure_encoding(Encoding::UTF_8 :external_encoding => Encoding::UTF_8, :invalid_characters => :drop )
Assume we have a legacy database and some of the fields contain Shift JIS, while some of the newer fields contain UTF-8 because someone screwed up the server configuration. You’re not even sure the encoding property on the strings you got make any sense because your database adapter is confused too.
example.ensure_encoding('UTF-8', :external_encoding => [Encoding::Shift_JIS, Encoding::UTF_8], :invalid_characters => :transcode )
As a last resort you’re trying to read some random files from disk and you have no idea what the external encoding is. You’ve just read them as binary and are hoping to make some sense from the data.
example.ensure_encoding('UTF-8', :external_encoding => :sniff, :invalid_characters => :transcode )
Note that the encoding sniffer is currently very naive and might not always be of any help.