I made a Spring REST application where you can perform CRUD operations based on HTTP methods of POST, PUT, GET, DELETE. I have the typical URI template of
http://host/root/{id}/{name}/{address} and etc.
We have a client who is accessing this REST service. Apparently they are sending parameters for multi-word name and address in the following form:
http://host/root/11/John+Smith/10+Las+Vegas+USA
They are using the HTML encoding scheme based on application/x-www-form-urlencoded type. According to the article in Wikipedia
The application/x-www-form-urlencoded
type
The encoding used by default is based
on a very early version of the general
URI percent-encoding rules, with a
number of modifications such as
newline normalization and replacing
spaces with "+" instead of "%20". -
http://en.wikipedia.org/wiki/Percent-encoding
However it appears the standard URL encoding scheme is to use %20 in replacing spaces in URI templates. Which one is correct?
My Spring REST automatically converts %20 to spaces. It's interpreted correctly. I'm using Spring 3.0.4. When + is met by my REST service, it's accepted as is. Of course when I put validation to exclude +, it is indeed excluded as expected.
Am I within standards or are there such double standards? Or is the client using an ancient scheme?
The point is that application/x-www-form-urlencoded can be used only in request parameters, whereas percent encoding is also supported in a path.
So,
http://host/root/11/?name=John+Smith&address=10+Las+Vegas+USA
is fine and will be properly decoded by Spring MVC, but
http://host/root/11/John+Smith/10+Las+Vegas+USA
is wrong and Spring MVC doesn't decode it, because the following form should be used instead:
http://host/root/11/John%20Smith/10%20Las%20Vegas%20USA
Related
We are exposing swagger-ui from our webserver using the Swashbuckle package working on top of Asp.Net Core. We are hitting an input validation issue for our Guid input fields.
The GUIDs we are pasting in, which are read from other parts of the system, are formatted as deb83f8a3edc4b78a2ece3321f81b58b, note the missing dashes. The input validation rejects this as it expects dashes in the format (so it accepts deb83f8a-3edc-4b78-a2ec-e3321f81b58b).
The swagger document that we serve has the parameter as type: string and format: uuid. It calls to some internal validationGuid call that have a reg-ex that forces the dashes. From the browser Console it seems like it is looking for a component called JsonSchema_string_uuid but is not finding it.
So my question is how can I extend swagger-ui to override the validation of specific parameter type/formats?
UPDATE:
I was made aware of the RFC that specifies a UUID as containing dashes and a workaround. However, I'm still interested in learning about ways to customize the validation of both custom formats and specifically uuid?
While the OpenAPI Specification and JSON Schema do not currently define format: uuid, RFC 4122 defines UUID as containing dashes, and some comments in the OpenAPI repository suggest that format: uuid is supposed to follow RFC 4122. This means your example without dashes is most likely not format: uuid.
Consider replacing format: uuid with pattern: '^[A-Fa-f0-9]{32}$'.
I have a Spring application that receives a request like http://localhost/foo?email=foo+bar#example.com. This triggers a controller that roughly looks like this:
#RestController
#RequestMapping("/foo")
public class FooController extends Controller {
#GetMapping
public void foo(#RequestParam("email") String email) {
System.out.println(email)
}
}
By the time I can access email, it's been converted to foo bar#example.com instead of the original foo+bar#example.com. According to When to encode space to plus (+) or %20? this should only happen in requests where the content is application/x-www-form-urlencoded. My request has a content type of application/json. The full MIME headers of the request look like this:
=== MimeHeaders ===
accept = application/json
content-type = application/json
user-agent = Dashman Configurator/0.0.0-dev
content-length = 0
host = localhost:8080
connection = keep-alive
Why is Spring then decoding the plus as a space? And if this is the way it should work, why isn't it encoding pluses as %2B when making requests?
I found this bug report about it: https://jira.spring.io/browse/SPR-6291 which may imply that this is fixed on version 3.0.5 and I'm using Spring > 5.0.0. It is possible that I may misinterpreting something about the bug report.
I also found this discussion about RestTemplate treatment of these values: https://jira.spring.io/browse/SPR-5516 (my client is using RestTemplate).
So, my questions are, why is Spring doing this? How can I disable it? Should I disable it or should I encode pluses on the client, even if the requests are json?
Just to clarify, I'm not using neither HTML nor JavaScript anywhere here. There's a Spring Rest Controller and the client is Spring's RestTemplate with UriTemplate or UriComponentsBuilder, neither of which encode the plus sign the way Spring decodes it.
Original Answer
You are mixing 2 things, a + in the body of the request would mean a space when header has application/x-www-form-urlencoded. The body or content of the request would be dependent on the headers but a request can just have a url and no headers and no body.
So the encoding of a URI cannot be controlled by any headers as such
See the URL Encoding section in https://en.wikipedia.org/wiki/Query_string
Some characters cannot be part of a URL (for example, the space) and some other characters have a special meaning in a URL: for example, the character # can be used to further specify a subsection (or fragment) of a document. In HTML forms, the character = is used to separate a name from a value. The URI generic syntax uses URL encoding to deal with this problem, while HTML forms make some additional substitutions rather than applying percent encoding for all such characters. SPACE is encoded as '+' or "%20".[10]
HTML 5 specifies the following transformation for submitting HTML forms with the "get" method to a web server.1 The following is a brief summary of the algorithm:
Characters that cannot be converted to the correct charset are replaced with HTML numeric character references[11]
SPACE is encoded as '+' or '%20'
Letters (A–Z and a–z), numbers (0–9) and the characters '*','-','.' and '_' are left as-is
All other characters are encoded as %HH hex representation with any non-ASCII characters first encoded as UTF-8 (or other specified encoding)
The octet corresponding to the tilde ("~") is permitted in query strings by RFC3986 but required to be percent-encoded in HTML forms to "%7E".
The encoding of SPACE as '+' and the selection of "as-is" characters distinguishes this encoding from RFC 3986.
And you can see the same behaviour on google.com as well from below screenshots
Also you can see the same behaviour in other frameworks as well. Below is an example of Python Flask
So what you are seeing is correct, you are just comparing it with a document which refers to the body content of a request and not the URL
Edit-1: 22nd May
After debugging it seems the decoding doesn't even happen in Spring. I happens in package org.apache.tomcat.util.buf; and the UDecoder class
/**
* URLDecode, will modify the source.
* #param mb The URL encoded bytes
* #param query <code>true</code> if this is a query string
* #throws IOException Invalid %xx URL encoding
*/
public void convert( ByteChunk mb, boolean query )
throws IOException
{
int start=mb.getOffset();
And below is where the conversion stuff actually happens
if( buff[ j ] == '+' && query) {
buff[idx]= (byte)' ' ;
} else if( buff[ j ] != '%' ) {
This means that it is an embedded tomcat server which does this translation and spring doesn't even participate in this. There is no config to change this behaviour as seen in the class code. So you have to live with it
SPR-6291 fixed this problem in v3.0.5 but this remains unresolved in some other cases like SPR-11047 is still unresolved. While SPR-6291's priority was Major, SPR-11047's priority is Minor.
I faced this problem when I was working on REST API in old Spring last year. There are multiple ways we can get data in Spring controller. So two of them are via #RequestParam or #PathVariable annotation
As others mentioned I think its spring's internal issue and does not specifically belong to URL encoding because I was sending data over POST request but it is somewhat encoding problem. But I also agree with others as now it remains problematic only in URL.
So there are two solutions I know:
You can use #PathVariable instead of #RequestParam because as of SPR-6291 this plus sign issue is fixed in #PathVariable and still remains open for #RequestParam as SPR-11047
My version of spring was not even accepting plus sign via #PathVariable annotation, so this is how I overcome the problem (I don't remember it step by step but it will give you hint).
In your case you can get the fields via JS and escape the plus sign before sending a request. Something like this:
var email = document.getElementById("emailField").value;
email = email.replace('+', '%2B');
If you have this request:
http://localhost/foo?email=foo+bar#example.com
then the original is foo bar#example.com. If you say the original should be foo+bar#example.com then the request should be:
http://localhost/foo?email=foo%2Bbar#example.com
So Spring is working as supposed to. Maybe on client you should check if the URI is properly encoded. The client-side URL encoding is responsible for building a correct HTTP request.
See encodeURI() if you generate the request in JavaScript or uriToString() if you generate the request in Spring.
Build your request string (the part after ?), without any encoding, with unencoded values like foo+bar#email.com, and only in the end, before actually using it in GET, encode all of it with whatever is available on the client platform. If you want to use POST then you should encode it according to the MIME type of your choice.
I'm trying to use Spring's ServletUriComponentsBuilder to create paging next and prev links from the current request.
The problem I'm having is that the ServletUriComponentsBuilder.fromCurrentRequest() is not unencoding percent-encoded values like:
http://example.com/articles?page%5Bnumber%5D=2
The problem is uses could have called the page with unencoded square brackets like http://example.com/articles?page[number]=2 without any problems.
Spring Data is accepting both variants (both unencoded square brackets and encoded square brackets) in it's pageable argument resolver.
This to the fact that under water the Coyote web request get parameter is used which contains the unencoded param names.
Also Spring's #RequestParam("page[number]") accepts without any problem the encoded request like http://example.com/articles?page%5Bnumber%5D=2.
From the server side I always want to return percent encoded url's as per RFC 3986.
But there does not seem a way to this as the UriComponents query params might contain both encoded en uncoded names. Because to that, if I would call encode() on the builder the already encoded query params get encoded another time, but if would contain unencoded names toURI() will fail as an unencoded [ is not allowed.
Note that the url's might contain multiple query params besides paging, e.g. for filtering.
A request could come in like:
http://example.com/articles?filter[category]=food
And would return a response with a encoded next link like:
http://example.com/articles?page%5Bnumber%5D=2&filter%5Bcategory%5D=food
My workaround it to ignore ServletUriComponentsBuilder and simply get the request url and do so custom regexp replacing.
I know this is an older question and you also found a workaround. But did you try to use ServletUriComponentsBuilder's build() method?
Some kind of the following:
ServletUriComponentsBuilder.fromCurrentRequest().build().toUriString();
I had some issues when handling JSON Strings and this helped.
I have an AngularJS Frontend and a Spring MVC Backend with Jackson to take care of the Serialization and JS<->Java conversion
When i pass German characters like "ö, ä, ü, ß" to my backend via http body payload, there is no problem. I have the header "Content-Type" "application/json;charset=UTF-8" and all works fine.
But if i have those characters in my url angular encodes them. This is fine however it encodes them a different way that jackson tries to decode i believe.
Here is what Angular makes out of "höhe": h%C3%B6he
I believe Jackson expects: h%f6he
I think this is because UTF8 is 2 byte while ASCII is 1 byte encoding. However is there a setting for either Jackson or Angular to "speak the same encoding language"?
Thanks for any help!
Kind regards,
Pascal
Jackson does not handle URL decoding, as it requires an input source such as InputStream or String: it is most likely that the Servlet container (Jetty?) that service runs on handles this. One problem is that definition of which encoding URL should use is... well, poorly defined really: "Content-Type" does NOT define this (it's just for payload).
So you need to figure out how to make servlet container and client have shared understanding of what encoding is to be used (difference in your case looks like UTF-8 vs Latin-1).
Or: if you can make client escape all non-ASCII characters with JSON escape sequences, that will also work.
I am using the Spring RestTemplate to make calls to a Apache Solr index. I form a request string manually and don't supply any intentional {variable_name} template expansion variables. Part of the query is the term {!lucene q.op=OR}. Unfortunately this gets processed by the URITemplate engine as part of a restTemplate.getForObject call.
Ideally i would like to stop this processing. Is there away of escaping the { } characters so that URITemplate doesn't process them? I have tried encoding the characters but RestTemplate assumes a non-encoded string so they are encoded twice and cause a 400: Bad Request on the backend.
Sample URL:
http://localhost/solr/select?q={!lucene
q.op=OR}se_genbanklocus:*
se_gb_create:* se_gb_update:*
se_clone_name:*
se_sample_tissue:*&facet=true&facet.limit=3&facet.mincount=1&facet.field=se_sample_tissue&facet.field=se_sample_tissue_name&facet.field=se_sample_tissue_code&facet.field=se_sample_tissue_class&facet.field=se_nuc_acid_type&facet.field=ssam_sample_georegion&start=0&rows=10
I've found a work around in which i can use the template to expand one variable which contains the offending {!lucene q.op=OR}
restTemplate.getForObject(solrServer+"select?{query}" , String.class, requestString );
The problem here is that you're using RestTemplate for something it's not designed for. The sample URL you gave is not a REST-style URL, it's just a mass of query parameters, using encoded characters that you're not going to find in a REST scheme, hence the difficulty with unwanted substitutions.
How about using the overloaded method that accepts a URI?