Reactive RedirectView throws exception if no RequestDataValueProcessor is present [SPR-15136]

[spring-projects-issues]

Daniel Fernández opened SPR-15136 and commented

The current reactive implementation of RedirectView makes the assumption that the application context will always contain an implementation of RequestDataValueProcessor, which does not seem to be true. Actually it is not RedirectView to be blamed but its parent AbstractView. But using RedirectView seems to be the most common way of being hit by this issue.

The RedirectView#createUrl(...) method does this just after building the URL and before returning it:

RequestDataValueProcessor processor = getRequestDataValueProcessor();
return (processor != null ? processor.processUrl(exchange, result) : result);

See it here on Github

Given RedirectView extends from AbstractView, this calls the AbsctractView#getRequestDataValueProcessor() method, which looks like this:

protected RequestDataValueProcessor getRequestDataValueProcessor() {
	if (getApplicationContext() != null) {
		String beanName = REQUEST_DATA_VALUE_PROCESSOR_BEAN_NAME;
		return getApplicationContext().getBean(beanName, RequestDataValueProcessor.class);
	}
	return null;
}

See it here on GitHub

Note that the code above unconditionally calls ApplicationContext#getBean(String,Class<?>), which will throw NoSuchBeanDefinitionException if there is no RequestDataValueProcessor registered.

In contrast, see how this is done at the MVC side, at the non-reactive RedirectView#updateTargetUrl(...) method, which calls the RequestDataValueProcessor only if it exists in the application context.

if (wac != null && wac.containsBean(RequestContextUtils.REQUEST_DATA_VALUE_PROCESSOR_BEAN_NAME)) {
	RequestDataValueProcessor processor = wac.getBean(
			RequestContextUtils.REQUEST_DATA_VALUE_PROCESSOR_BEAN_NAME, RequestDataValueProcessor.class);
	return processor.processUrl(request, targetUrl);
}

See it here on GitHub.

---

Affects: 5.0 M4

Referenced from: commits 8b7f3a6

[spring-projects-issues]

Daniel Fernández commented

I've just submitted a Pull Request fixing this here: #1296