[
https://issues.apache.org/jira/browse/SLING-8279?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mark Adamcin updated SLING-8279:
--------------------------------
Description:
With Sling Models, it is very easy to construct composite model types from more
than one resource, usually addressable as a subtree of the repository.
However, a pattern is emerging where mapping is being performed to create
related links within the model, which mandates that SlingHttpServletRequest be
used as the adaptable type, because a Resource adaptable would not provide
sufficient context for ResourceResolver.map(). Without a request for context,
.map() would likely return incorrect links based on a default base request URL
of "http://localhost/";.
While on one hand, it might be argued that link mapping should only occur in
the view, after the model graph has been constructed, on the other hand, link
resolution and resource mapping are so fundamental to the correct behavior of a
web application that we should take advantage of every opportunity to make
these activities more convenient and less error prone.
If you trace the code for resource resolution and mapping, you will find that
it relies on just four discrete contextual properties that are currently
available only from a request object (i.e. not available from a Resource or its
ResourceMetadata):
# scheme
# host
# port
# contextPath
In addition, given that the ResourceResolver used by servlets when handling a
request is generally retrieved from the Sling Request itself using
getResourceResolver(), it seems redundant in concept, not to mention clumsy in
practice, to require passing the request as an argument back to the resource
resolver (that was created specifically for the request in question) in order
to render links for any resources resolved while servicing that request.
I think it is time to change the expected behavior of
ResourceResolver.resolve(String), ResourceResolver.map(String path), and other
ResourceResolver methods that return resources without an explicit
HttpServletRequest parameter, such that:
# request.getResourceResolver().resolve(path) returns the same Resource as
(any ResourceResolver).resolve(request, path)
# request.getResourceResolver().map(path) returns the same String as (any
ResourceResolver).map(request, path)
#
request.getResourceResolver().getResource(somePath).getResourceResolver().resolve(path)
returns the same Resource as request.getResourceResolver().resolve(path)
#
request.getResourceResolver().findResources(someQuery).next().getResourceResolver().resolve(path)
returns the same Resource as request.getResourceResolver().resolve(path)
# etc.
# ResourceResolverFactory.getResourceResolver(Map) and
ResourceResolverFactory.getServiceResourceResolver(Map) would return
ResourceResolvers that continue to use [http://localhost:80|http://localhost/]
as the default context url.
If these constraints can not be satisfied reasonably using the existing
resolve(String) and map(String) methods, I would propose adding overloads that
accept a context Resource in place of the context HttpServletRequest, with
additional properties added to ResourceMetadata during request resource
resolution that persist the four request context properties listed above
(scheme, host, port, contextPath).
was:
With Sling Models, it is very easy to construct composite model types from more
than one resource, usually addressable as a subtree of the repository.
However, a pattern is emerging where mapping is being performed to create
related links within the model, which mandates that SlingHttpServletRequest be
used as the adaptable type, because a Resource adaptable would not provide
sufficient context for ResourceResolver.map(). Without a request for context,
.map() would likely return incorrect links based on a default base request URL
of "http://localhost/";.
While on one hand, it might be argued that link mapping should only occur in
the view, after the model graph has been constructed, on the other hand, link
resolution and resource mapping are so fundamental to the correct behavior of a
web application that we should take advantage of every opportunity to make
these activities more convenient and less error prone.
If you trace the code for resource resolution and mapping, you will find that
it relies on just four discrete contextual properties that are currently
available only from a request object (i.e. not available from a Resource or its
ResourceMetadata):
# scheme
# host
# port
# contextPath
In addition, given that the ResourceResolver used by servlets when handling a
request is generally retrieved from the Sling Request itself using
getResourceResolver(), it seems redundant in concept, not to mention clumsy in
practice, to require passing the request as an argument back to the resource
resolver (that was created specifically for the request in question) in order
to render links for any resources resolved while servicing that request.
I think it is time to change the expected behavior of
ResourceResolver.resolve(String), ResourceResolver.map(String path), and other
ResourceResolver methods that return resources without an explicit
HttpServletRequest parameter, such that:
# request.getResourceResolver().resolve(path) returns the same Resource as
(any ResourceResolver).resolve(request, path)
# request.getResourceResolver().map(path) returns the same String as (any
ResourceResolver).map(request, path)
#
request.getResourceResolver().getResource(somePath).getResourceResolver().resolve(path)
returns the same Resource as request.getResourceResolver().resolve(path)
#
request.getResourceResolver().findResources(someQuery).next().getResourceResolver().resolve(path)
returns the same Resource as request.getResourceResolver().resolve(path)
# etc.
If these constraints can not be satisfied reasonably using the existing
resolve(String) and map(String) methods, I would propose adding overloads that
accept a context Resource in place of the context HttpServletRequest, with
additional properties added to ResourceMetadata during request resource
resolution that persist the four request context properties listed above
(scheme, host, port, contextPath).
> Having a Resource + ResourceMetadata should be sufficient for roundtrip link
> mapping.
> -------------------------------------------------------------------------------------
>
> Key: SLING-8279
> URL: https://issues.apache.org/jira/browse/SLING-8279
> Project: Sling
> Issue Type: Wish
> Components: ResourceResolver
> Reporter: Mark Adamcin
> Priority: Major
>
> With Sling Models, it is very easy to construct composite model types from
> more than one resource, usually addressable as a subtree of the repository.
> However, a pattern is emerging where mapping is being performed to create
> related links within the model, which mandates that SlingHttpServletRequest
> be used as the adaptable type, because a Resource adaptable would not provide
> sufficient context for ResourceResolver.map(). Without a request for context,
> .map() would likely return incorrect links based on a default base request
> URL of "http://localhost/";.
> While on one hand, it might be argued that link mapping should only occur in
> the view, after the model graph has been constructed, on the other hand, link
> resolution and resource mapping are so fundamental to the correct behavior of
> a web application that we should take advantage of every opportunity to make
> these activities more convenient and less error prone.
> If you trace the code for resource resolution and mapping, you will find that
> it relies on just four discrete contextual properties that are currently
> available only from a request object (i.e. not available from a Resource or
> its ResourceMetadata):
> # scheme
> # host
> # port
> # contextPath
> In addition, given that the ResourceResolver used by servlets when handling a
> request is generally retrieved from the Sling Request itself using
> getResourceResolver(), it seems redundant in concept, not to mention clumsy
> in practice, to require passing the request as an argument back to the
> resource resolver (that was created specifically for the request in question)
> in order to render links for any resources resolved while servicing that
> request.
> I think it is time to change the expected behavior of
> ResourceResolver.resolve(String), ResourceResolver.map(String path), and
> other ResourceResolver methods that return resources without an explicit
> HttpServletRequest parameter, such that:
> # request.getResourceResolver().resolve(path) returns the same Resource as
> (any ResourceResolver).resolve(request, path)
> # request.getResourceResolver().map(path) returns the same String as (any
> ResourceResolver).map(request, path)
> #
> request.getResourceResolver().getResource(somePath).getResourceResolver().resolve(path)
> returns the same Resource as request.getResourceResolver().resolve(path)
> #
> request.getResourceResolver().findResources(someQuery).next().getResourceResolver().resolve(path)
> returns the same Resource as request.getResourceResolver().resolve(path)
> # etc.
> # ResourceResolverFactory.getResourceResolver(Map) and
> ResourceResolverFactory.getServiceResourceResolver(Map) would return
> ResourceResolvers that continue to use
> [http://localhost:80|http://localhost/] as the default context url.
> If these constraints can not be satisfied reasonably using the existing
> resolve(String) and map(String) methods, I would propose adding overloads
> that accept a context Resource in place of the context HttpServletRequest,
> with additional properties added to ResourceMetadata during request resource
> resolution that persist the four request context properties listed above
> (scheme, host, port, contextPath).
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
