Spring URIEditor double escapes % signs for URIs that contain a schema [SPR-6005]

Andreas Hartl opened SPR-6005 and commented

org.springframework.beans.propertyeditors.URIEditor does a double escaping of URLs with a schema that use % escaping.

Test:

System.out.println(new SimpleTypeConverter().convertIfNecessary("http://de.wikipedia.org/wiki/%C3%96", URI.class));

prints

http://de.wikipedia.org/wiki/%25C3%2596

I ran into this behavior when I wanted to make use of URIEditor's feature that it can resolve classpath: URLs. If the classpath contains spaces or other characters that need to be escaped in URLs, then the resulting URI contains double escaped % signs.

My current workaround is to extend URIEditor and override createURI to always return new URI(value). when registered as custom editor for URIs, this PropertyEditor returns the proper URI.

Affects: 3.0 M3

Attachments: - SPR-6005-URIEditor-patch.txt (2.44 kB)

Issue Links: - #21123 URIEditor should not double escape classpath: URIs

Referenced from: commits https://github.com/spring-projects/spring-framework/commit/7ec9f1506a3d93a5dacc66c00aec5ca6943a5ae7

Comment From: spring-projects-issues

Edison Figueroa commented

Currently the URIEditor relies on the java.net.URI, specifically using the following constructor for URI prefixed with a schema:

URI(String scheme, String ssp, String fragment)

From the Java API:


This constructor first builds a URI in string form using the given components as follows: 

   1. Initially, the result string is empty.
   2. If a scheme is given then it is appended to the result, followed by a colon character (':').
   3. If a scheme-specific part is given then it is appended. *Any character that is not a legal URI character is quoted*.
   4. Finally, if a fragment is given then a hash character ('#') is appended to the string, followed by the fragment. Any character that is not a legal URI character is quoted.

By using this constructor the URIEditor assumes that the provided URI might contain illegal URI characters, and always applies URI character encoding.

Alternatively, the editor could accommodate for already encoded URIs. This could be accomplished by initially searching for escaped characters in the source URI, and if present then simply call the static create(String str) method.

I've attached a patch to the URIEditor and URIEditorTests

Comment From: spring-projects-issues

Arjen Poutsma commented

Detecting whether a given URI is encoded or not is a tricky path. Basically, there is no way to know whether a % followed by two hexadecimals signifies an escape sequence, or is just a % with two characters behind it.

For instance, the String http://example.com/foo%20bar could either mean http://example.com/foo bar, or http://example.com/foo%2520bar; there is no way to tell.

So rather than detecting encoding, I'm going to add a boolean property to the URLEditor that indicates whether the editor should encode the given string or not.