I'm making a cross-platform application that renames files based on data retrieved online. I'd like to sanitize the Strings I took from a web API for the current platform.
I know that different platforms have different file-name requirements, so I was wondering if there's a cross-platform way to do this?
Edit: On Windows platforms you cannot have a question mark '?' in a file name, whereas in Linux, you can. The file names may contain such characters and I would like for the platforms that support those characters to keep them, but otherwise, strip them out.
Also, I would prefer a standard Java solution that doesn't require third-party libraries.
It is not clear from your question, but since you are planning to accept pathnames from a web form (?) you probably ought block attempts renaming certain things; e.g. "C:\Program Files". This implies that you need to canonicalize the pathnames to eliminate "." and ".." before you make your access checks.
Given that, I wouldn't attempt to remove illegal characters. Instead, I'd use "new File(str).getCanonicalFile()" to produce the canonical paths, next check that they satisfy your sandboxing restrictions, and finally use "File.exists()", "File.isFile()", etc to check that the source and destination are kosher, and are not the same file system object. I'd deal with illegal characters by attempting to do the operations and catching the exceptions.
or just do this:
Result:
A20_B22b_A_BC_ld_ma.la.xps
Explanation:
[a-zA-Z0-9\\._]
matches a letter from a-z lower or uppercase, numbers, dots and underscores[^a-zA-Z0-9\\._]
is the inverse. i.e. all characters which do not match the first expression[^a-zA-Z0-9\\._]+
is a sequence of characters which do not match the first expressionSo every sequence of characters which does not consist of characters from a-z, 0-9 or . _ will be replaced.
There's a pretty good built-in Java solution - Character.isXxx().
Try
Character.isJavaIdentifierPart(c)
:Result is "name.é$_".
Paths.get(...)
throws a detailed exception with the position of the illegal character.This is based on the accepted answer by Sarel Botha which works fine as long as you don't encounter any characters outside of the Basic Multilingual Plane. If you need full Unicode support (and who doesn't?) use this code instead which is Unicode safe:
Key changes here:
length
instead of justlength
charAt
append
char
s toint
s. In fact, you should never deal withchar
s as they are basically broken for anything outside the BMP.As suggested elsewhere, this is not usually what you want to do. It is usually best to create a temporary file using a secure method such as File.createTempFile().
You should not do this with a whitelist and only keep 'good' characters. If the file is made up of only Chinese characters then you will strip everything out of it. We can't use a whitelist for this reason, we have to use a blacklist.
Linux pretty much allows anything which can be a real pain. I would just limit Linux to the same list that you limit Windows to so you save yourself headaches in the future.
Using this C# snippet on Windows I produced a list of characters that are not valid on Windows. There are quite a few more characters in this list than you may think (41) so I wouldn't recommend trying to create your own list.
Here is a simple Java class which 'cleans' a file name.
EDIT: As Stephen suggested you probably also should verify that these file accesses only occur within the directory you allow.
The following answer has sample code for establishing a custom security context in Java and then executing code in that 'sandbox'.
How do you create a secure JEXL (scripting) sandbox?