Aryeh,
On 9/12/23 17:50, Aryeh Friedman wrote:
On Tue, Sep 12, 2023 at 1:51 PM Christopher Schultz
<ch...@christopherschultz.net> wrote:
Aryeh,
On 9/12/23 12:42, Aryeh Friedman wrote:
On Tue, Sep 12, 2023 at 11:42 AM Christopher Schultz
<ch...@christopherschultz.net> wrote:
Aryeh,
On 9/11/23 10:05, Aryeh Friedman wrote:
On Mon, Sep 11, 2023 at 9:47 AM Christopher Schultz
<ch...@christopherschultz.net> wrote:
Aryeh,
On 9/9/23 19:36, Aryeh Friedman wrote:
On Sat, Sep 9, 2023 at 1:23 PM Mark Thomas <ma...@apache.org> wrote:
On 09/09/2023 11:52, Aryeh Friedman wrote:
Every other jsp in my webapp (and other webapps on the same tomcat
instance [9.0.75]) works and I am using a the default container but as
curl/catalina.out show BasePage is *NEVER* being called (either the
_jspService() or the getX()):
How have you configured your JSP(s) to use this alternative base class?
sudo cat /usr/local/apache-tomcat-9.0/webapps/tlaitc-dashboard-1a1/index.jsp
<!-- Copyright (C) 2023 TLAITC and contributors -->
<%@page extends="dashboard.web.pages.BasePage"%>
hi x is ${x}
Output shown in log (sorry for not including the JSP the first time)
but to make it easier to find the output is "hi x is " when it should
be "hi x is 123234"... as you notice there are zero errors/warning in
catalina but there is none of the println's also... so the only thing
I can surmise is BasePage is never being called <%@page
extends="dashboard.web.pages.BasePage"%> somehow failed but I have
verified that correct spelling several times and also verified any
syntextual errors [including the contents of the string literal] will
show up in catalina.out (i.e. wrong class name is logged as an error)
Your _jspService method in your base class will never be called, because
it is overridden by the auto-generated class for your JSP, which does
not call super._jspService.
I do not believe that this:
Hi X is ${x}
...will result in this.getX() being called from the JSP. References in
EL ${...} expressions will be resolved against the PageContext (and
other wider contexts), not against the JSP class currently executing.
If you want to call this.getX() then you will need to do this:
Hi X is <% getX() %>
I wouldn't bother messing-around with class hierarchies in JSP. It
usually doesn't get you much but almost always requires you to bind
yourself very closely with the specific implementation of the JSP engine.
It would be far better to use typical MVC-style development where a
servlet (or similar) handles the real work of the request, possibly
including writing a value of "x" to the request attributes. Then forward
your request to your JSP to produce the response content. This will be
much more straightforward and you will have fewer problems like this,
where expected behavior is confused by all the magic that JSP provides.
Thanks but I have a very specific use case which the following working
example below should make more clear:
<!-- Copyright (C) 2023 TLAITC and contributors -->
<%@page import="dashboard.web.page.Page"%>
<%
// THIS WOULD NOT BE NEEDED if <%@page extends="..."%> worked
//
// for now we don't need to keep the page object just
// the setAttributes in our ctx
new Page(pageContext);
%>
<html>
<head>
<%@include file="/widgets/scripts/scripts.jsp"%>
</head>
<body>
<jsp:include page="${pagePath}"/>
</body>
</html>
and the Page class:
package dashboard.web.page;
import javax.servlet.http.HttpServletRequest;
import javax.servlet.jsp.PageContext;
import org.petitecloud.util.PetiteCloudNullException;
// making this extend the right subclass plus working page
extends="...." would mean zero in-line Java
// Copyright (C) 2023 TLAITC and contributors
public class Page
{
public Page(PageContext ctx)
{
_dbc_construction(ctx);
this.ctx=ctx;
HttpServletRequest req=(HttpServletRequest) ctx.getRequest();
String[] parts=req.getRequestURI().split("/");
int split=2;
if(parts[0].equals("http:"))
split+=2;
name="";
for(int i=split;i<parts.length;i++)
name+="/"+parts[i];
if(name.length()==0)
name="/index.jsp";
// we can safely asssume all valid requests will end with
// .jsp at this point
name=name.substring(1,name.length()-".jsp".length());
path="/content/"+name+"/main.jsp";
ctx.setAttribute("pagePath",path);
}
// only used in testing
public Page(String name)
{
this.name=name;
}
public String getName()
{
return name;
}
public String getPath()
{
return path;
}
public PageContext getPageContext()
{
return ctx;
}
@Override
public String toString()
{
return name;
}
@Override
public int hashCode()
{
return name.hashCode();
}
@Override
public boolean equals(Object o)
{
if(o==null||!o.getClass().equals(getClass()))
return false;
Page other=(Page) o;
return name.equals(other.name);
}
private String name;
private String path;
private PageContext ctx;
// ----- # DBC # -------------------------------------------
private void _dbc_construction(PageContext ctx)
{
if(ctx==null)
throw new PetiteCloudNullException("ctx is null");
}
}
So all I am attempting to do is write it like this (get rid of the
last remaining in-line java but not be forced to use tag instead of
${} either):
<!-- Copyright (C) 2023 TLAITC and contributors -->
<%@page extends="dashboard.web.page.Page"%>
<html>
<head>
<%@include file="/widgets/scripts/scripts.jsp"%>
</head>
<body>
<jsp:include page="${pagePath}"/>
</body>
</html>
Side note I am currently adding user detection to the above page class
to that it can auto-enforce ACL's
I'm not sure I understand the point of the whole exercise. Nothing you
have in the PageContext class constructor cannot be done from within the
JSP itself, or -- better yet -- in an <%include> used by as many pages
as you want. *OR* ... you could do it in a Filter and apply it to all
requests, storing the information in the request attributes, which is
much more standard than directly modifying the page context.
The idea is to avoid any custom entries in web.xml (which filters
require)
This is not entirely true, and seems to be an arbitrary requirement. The
application is configured via web.xml. Why disallow any configuration in
web.xml?
Because of the need to run it on a potentially stripped down
(embedded) non servlet based (but Java) based HTTP service/servlet
(whether you want to admit it or not HTTP is used for a lot more then
just HTML and web browsers). Many such servers are hard wired to
their application and thus the code would not be reusable with any
page framework they might use (the one I have in mind does use
sessions but they are made controlled by the caller).
Hmm. Okay.
and the entire point is this is a top level template and
future versions will provide a number of standard setAttriburtes that
expose different things like the logged in user.
Yeah, you are welcome to your own opinion, but everything you describe
sounds like "we want to use JSPs for everything and not write Java code"
and honestly what you need is plain-old Java code for this stuff.
No, you read it completetly wrong we want 100% test coverage on any
code that can create dynamic results (java or JSP) and frankly without
making a fake set of servlet primitives (which we have done for just
this kind of testing) it is almost impossible to unit test stuff that
deals directly with the servlet/jsp-api's and thus we need to push as
much as absolutely possible out of un-testable compilation units/files
(aka no Java in JSP).... and the form of JSP that still keeps the
dynamics we want (i.e. it is visually obvious what is dynamic content
and what is static content from just reading it) is the ${...} form.
(Note this is also compatible with a batch processed mail-merge like
text macro expander we made for other purposes so 2 birds with the
same stone).
Just so you know the application is soft-life critical (remote cardiac
monitoring w/ webapp frontend) and is 100% Java and currently does the
standard intermix of in-line Java vs. non-in-line Java in JSP's but as
I said above due to the critically of the app we have decided to make
blank rule that all lines of Java must be unit tested if at all
possible (i.e. not insanely hard to trigger all paths).
So you have a requirement to test all Java code but not all JSP code?
JSP is just elaborate syntactic sugar for Java.
In general I try to avoid relying on any given low level framework
then is needed and doing this ties us more to servlet-api then makes
sense (i.e. it is not very portable to non-monolithic servers like
tomcat but it is portable if you avoid stuff like web.xml as
possible).
I would argue the /absolute opposite/: using web.xml for its intended
purpose is exactly the correct way to avoid tying yourself to a specific
server. If you are trying to avoid bindings to the Servlet API but in
exchange you are binding to Tomcat, I see that as a net-negative: the
Servlet API is a better choice than Tomcat if you have to pick one to
bind to.
There are frameworks which you can use that do not require you do bind
your application to either Tomcat OR the servlet-api (I'm thinking of
Apache Struts 2, for example), but you do have to give up some small
bits of capabilities because really advanced stuff requires binding to
those APIs.
Again you missed the point the idea is to make it so it can run on a
servlet based engine or non-servlet based one and the non-servlet
based one would need the configurations that are kept in web.xml (and
not have the built-in capacity to read it). Think reuse.
Additionally any non-platform (servlet vs. non-servlet) code would be
in the servlets or other things.... I.e. regardless of the html/json
rendering engine used.
I think you lost me again. Non-servlet code goes in the servlets?
Writing JSPs is already very very VERY tightly binding your application
to the Servlet API. I don't understand this requirement at a very
fundamental level.
See above it is not tightly binding if you ${...} because many text
macro packages use that as a delim (it is almost the defacto
standard). Thus as long you present the same map of keys/values to
the ${...} it is 100% portable across many ecosystems where is
<%....%> is not for one reason it messes with PHP (which I don't use).
Oh.
You want to use JSP syntax with either a true JSP engine OR some other
technology which can produce the same kind of output? That seems fragile.
Why not just use something which has nothing to do with JSP at all?
Obvious candidates are Apache Velocity and FreeMarker. You can use both
of those without having to hand-wave your way around
servlet-vs-non-servlet. Notably, they also do not translate ultimately
into Java source.
The answer to the question "how do I write my application in a way that
avoids binding to (e.g) the servlet API" is "use a framework which
insulates you and the application from that API". There are some (again,
thinking of Apache Struts 2, here) that don't even require that you bind
your code to THEIR APIs. You write POJOs and some configuration to glue
it all together. No compile-time dependencies. You can port your
application to Swing/JavaFX if you feel like going down that route, and
only the GUI glue itself needs to be "bound" to
AWT/Swing/JavaFX/whatever. Your core application code doesn't need any
of that.
For security reasons (not depending on more 3rd party black boxes then
we have to) we use only the bare min libraries currently (servlet-api,
jsp-api, json, itext and the standard library, and keeping the list to
just those is purposeful.
The stuff the Servlet Spec provides should be thought of as glue that
allows your core application (without Servlet API bindings) to be
accessible through a web-based interface. But you appear to be trying to
implement your core application in JSP which is counter-intuitive to me
and seems to be the opposite of your goal to be API-independent.
Not counter unititive at all for reasons cited above (near defacto
standard text macro format).
Case in point for performance reasons I wrote a server called
babyHttp that is an extremely stripped down API endpoint oriented
server that does not use web.xml or anything of the kind (purposely
make it so the only linkage between modules is actual Java source
code). Keep in mind HIPAA is a requirement here.
I do not know why your "case in point" is relevant. Sure, you were able
to create a toy HTTP server but anybody who has read an introductory
Java programming book can probably copy such a thing out of the examples
from that same book. HIPAA is a red herring, other than you'd better be
careful about implementing your own HTTP endpoint since it's very easy
to do it incorrectly.
I am well aware of that having written about 6 or 7 HTTP servers since
the mid 90's including the first single threaded server, held the
record of most concurrent connections before 2000 at 10k concurrent
requests (that through a few forks turned into thttp). The only
reason for even bringing this up is I have learned over the years not
to question why someone has a specific use case (as long it is not
completely idiotic) and do my best to help them implement it the way
they see fit (aka domain experts know better then generalists)
Sounds like you know everything you need to know, then. I guess I'll
show myself out.
-chris
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
