Add support for prev/next/self links when returning paginated results. To do this: 1) change "data" envelope from an array to a dict 2) move the "data" array to the "collection" property, which is an array of elements in the collection. 3) add @links dict keyed by link relation: self, next, prev. Each relation is an array of dicts with uri and rel keys. In this case there is only one element, but there is nothing preventing a relation from having multiple url's. So this follows the formatting needed for the general case. Relations are present only if it makes sense. So first page has no prev and last page has no next. 4) add @total_size with number of element selected if they were not paginated. Replicates data in X-Count-Total header. Changed index to start at 1. So the first page is page_index 1 and not page_index 0. (So I am no longer surprised when I set page_index to 1 and am missing a bunch of records 8-)). Also a small fixup, json response ends with a newline so printing the data, or using curl makes sure that anything printing after the json output (like shell prompts) is on a new line. Tests added for all cases.

"""Exceptions for use across all Roundup components.
"""

__docformat__ = 'restructuredtext'

class LoginError(BaseException):
    pass

class Unauthorised(BaseException):
    pass

class Reject(BaseException):
    """An auditor may raise this exception when the current create or set
    operation should be stopped.

    It is up to the specific interface invoking the create or set to
    handle this exception sanely. For example:

    - mailgw will trap and ignore Reject for file attachments and messages
    - cgi will trap and present the exception in a nice format
    """
    pass


class RejectRaw(Reject):
    """
    Performs the same function as Reject, except HTML in the message is not
    escaped when displayed to the user.
    """
    pass


class UsageError(ValueError):
    pass

# vim: set filetype=python ts=4 sw=4 et si