* common.lisp (arguments-to-uri-1): Remove unreachable code.

Reported by Zach Beane.

* common.lisp (uri-to-arguments-1, arguments-to-uri-1): Delete
these two spurious and duplicate definitions.

Also fix ridiculous indentation.

* api.lisp (uri-to-arguments, arguments-to-uri)
(read-for-resource, write-for-resource, handle-request)
(documentation, backend-payload): Fix docstrings.

* common.lisp (find-content-class, verb-spec-or-lose)
(parse-resource, parse-content-type-header, check-optional-args)
(genpath-fn-lambda-list, explain-failsafe): Fix indentation.
(arguments-to-uri-1, uri-to-arguments-1, read-for-resource-1)
(write-for-resource-1): Rename from ARGUMENTS-TO-URI,
URI-TO-ARGUMENTS, READ-FOR-RESOURCE and WRITE-FOR-RESOURCE,
respectively.

* common.lisp (destructive-p): DELETE is also a destructive verb.
(handle-request-1): DELETE doesn't care about content-types

* snooze-tests.lisp (:snooze-tests): import some symbols from
snooze-tests-demo.
(:snooze-tests-demo): Move package definition to top.
(todo): Allow making TODO with given id.
(with-request): Bind SNOOZE:*CATCH-ERRORS* and SNOOZE:*CATCH-HTTP-CONDITIONS*
(todo :delete): New route
(test-delete-route): New test
(*mock-http-payload*): Move symbol to :SNOOZE-TESTS-DEMO

* common.lisp (resignalled-condition): Use a new slot
ORIGINAL-CONDITION-BACKTRACE.
(explain-failsafe): Improve text.
(explain-failsafe unconvertible-argument): New method.
(explain-failsafe :after): Print the "...when it was bitten by"
here. Explain the original condition and the backtrace.

* safe-simple-read.lisp (print-object :snooze-reader-error): New
method.
(safe-simple-read-from-string): Check for empty string.

Customizations to PRINT-OBJECT for conditions affect every attempt to
print these condition, which may lead to heap- and stack-exhausting
errors when printing backtraces, for example. This is what was
happening in issue #8.

Better to have a separate method to provide the more human readable
explanation and keep to using terse PRINT-UNREADABLE-OBJECT
expressions in the PRINT-OBJECT methods.

In the future, keep the original backtrace for the original condition
of the resignalled condition as a string.

* common.lisp (explain-condition-failsafe): Call a new internal
generic function EXPLAIN-FAILSAFE.
(resignalled-condition): Prepare to store backtrace of the
original condition.
(print-object http-condition): Use PRINT-UNREADABLE-OBJECT
(print-object :after resignalled-condition): Delete.
(print-object error-when-explaining, invalid-uri-structure)
(incompatible-lambda-list): Delete these methods.
(explain-failsafe): New methods to replace PRINT-OBJECT.

Thanks to 'lispm' on reddit.

* safe-simple-read.lisp (read-string, read-name, parse-symbol)
(safe-simple-read-from-string): Use EQL to compare chars.

* common.lisp (genpath-fn-lambda-list): New helper.
(make-genpath-form): Refactor slightly

SNOOZE-COMMON:PARSE-CONTENT-TYPE-HEADER presently ignores parameters
in the "Content-Type:" header value, such as "Charset=UTF-8", but at
least it sees the basic content type descriptor, whereas it previously
returned nil.

* common.lisp (content-classes-in-accept-string): Use
PARSE-CONTENT-TYPE-HEADER.
(parse-content-type-header): Use RFC2388:PARSE-HEADER.

* snooze.asd (:snooze): Use RFC2388 system.

The commit b1300e ("Resist more DOS keyword intern attacks") made the
code too strict when deciding whether to intern symbols used for
keyword arguments. This commit relaxes the fix, while still blocking
that particular DOS.

* common.lisp (uri-to-arguments): Be less strict when checking if
symbols naming keyword arguments should be interned in the
:KEYWORD package.

* api.lisp (*uri-content-types-function*): Silly typo.

This seriously needs to be reworked, hopefully without changing the
API.

* common.lisp (call-brutally-explaining-conditions): Handle case
where *CATCH-ERRORS* is nil but *CATCH-HTTP-CONDITIONS* is t.

SNOOZE:URI-TO-ARGUMENTS is more careful when interning the keys in
key-value pairs found in the URI as keyword arguments.

It will only do so when a corresponding symbol with the same name is
found in the resource's package. This heuristic should prevent abusive
keyword interns.

* common.lisp (resource-package): New helper.
(uri-to-arguments): Be more careful when interning.
(read-for-resource): Use RESOURCE-PACKAGE

* snooze-tests.lisp (test-parse-uri): Explicitly intern BAR before
testing.

* demo/lispdoc.lisp (cl-who:convert-tag-to-string-list): New hack.
(defform) Use *MACROEXPAND-HOOK* and
CL-WHO:CONVERT-TAG-TO-STRING-LIST.

* constants.lisp (*mime-type-list*): Add
"application/x-www-form-urlencoded".

* demo/lispdoc.lisp (#:snooze-demo): Import some hunchentoot symbols.
(defform): Idea for a DEFFORM macro.
(render-docstring-form): Refactor this out.
(lispdoc): New POST route
(redirect, explain-condition): New redirect condition and a
suitable explainer.

* api.lisp (make-hunchentoot-app): Use (funcall (fdefinition)))

* README.md: Improve section on READ-FOR-RESOURCE and
WRITE-FOR-RESOURCE.

* common.lisp (read-for-resource, write-for-resource): Better
docstrings. Write unqualified uninterned symbols.

* api.lisp (backend-payload): Move Clack and Hunchentoot
implementations to api.lisp for now.
(make-hunchentoot-app): New function.

* common.lisp (backend-payload): Moved to api.lisp

* package.lisp (#:snooze): Export MAKE-HUNCHENTOOT-APP.

* snooze.asd (:snooze-demo): Move demo to separate dir.

* common.lisp (explain-condition-failsafe): Use it.

* constants.lisp: New function REASON-FOR

* package.lisp (#:snooze-common): Export REASON-FOR

READER-ERROR doesn't take usual SIMPLE-ERROR args. (it does in Allegro though)

* safe-simple-read.lisp (snooze-reader-error): New error
condition.
(read-string, parse-symbol)
(safe-simple-read-from-string): Use it.

* common.lisp (uri-to-arguments): Don't resignal if *CATCH-ERRORS*
is nil. Fix indent.
(read-for-resource): Don't check for errors here and pass t to
SNOOZE-SAFE-SIMPLE-READ:SAFE-SIMPLE-READ-FROM-STRING.

* safe-simple-read.lisp (parse-symbol): Return 4 values.
(safe-simple-read-from-string): Take MAKE-SYMBOL-P. Update
docstring.

* snooze-tests.lisp (test-some-routes): Update test.
(safe-simple-read-should-error): Update tests.
(safe-simple-dont-read-back): Delete this test.

* api.lisp (read-for-resource): Update docstring.

* common.lisp (read-for-resource): Use
SNOOZE-SAFE-SIMPLE-READ:SAFE-SIMPLE-READ-FROM-STRING.

* package.lisp: New package.

* safe-simple-read.lisp: New file

* snooze-tests.lisp (safe-simple-read-back)
(safe-simple-dont-read-back, safe-simple-read-should-error): New tests

* snooze.asd (:snooze): Depends on PARSE-FLOAT. Add
"safe-simple-read".

* common.lisp (print-object): Move all PRINT-OBJECT for conditions nearer the conditions section.
(*resource*): Fix a silly bug binding *RESOURCE* to a docstring.
(default-resource-name, search-for-extension-content-type)
(all-defined-resources, backend-payload): Group together as
default implementation for options.
(uri-to-arguments, arguments-to-uri, read-for-resource)
(write-for-resource): Group together.

* snooze-tests.lisp (#:snooze-parse-uri-tests): Don't use this package.
(test-parse-uri, path-generation, path-parse-back): Adapt to new
default argument-reading rules.