Mailing-List: contact dev-help@apex.incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@apex.incubator.apache.org
MIME-Version: 1.0
Date: Tue, 1 Dec 2015 09:38:31 +0530
Message-ID: 
 <CAD0KA=O5ZfAnRv3PzuMC4oXaeXa+u2vGpOTDgxQXZoi0+xO7oA@mail.gmail.com>
Subject: [APEX-3/APEX-104] Module Separator.
From: Tushar Gosavi <tushar@datatorrent.com>
To: dev@apex.incubator.apache.org
Content-Type: multipart/alternative; boundary=001a114d7ec8910cda0525ce4ff6

--001a114d7ec8910cda0525ce4ff6
Content-Type: text/plain; charset=UTF-8

Hi All,

As part of module support, we will need to generate names for operators
internal to the module before adding it to the original DAG to avoid
conflict of names. For example if a developer adds a operator with name "A"
and module is name "M" in the DAG. Developer of 'M' is adding a operator
'A' into the module DAG. To avoid name conflict
with the operator in the main DAG we will generate name for internal
operators by concatenation of module name and operator name. The 'A' within
module 'M' will be added to main DAG as 'M_A'.  This works for recursive
modules too. (i.e module containing module).

Which separator should we use for module namespace? The requirement are
- Less chances of collision. (User can add operator with name 'M_A' too).
- Compatible with json format. (Should not have a problem while parsing
json response from rest api)
- Should be able to specify property in dt-site.xml.

While '_' works, In the pull request it has been pointed out that it has
higher chance of collision. Few choices are

1) _ (Underscore)
2) __ (double underscore)
3) : (colon, I will check compatibility with rest API, as it is used to
separate key and value in json)
4) :: (Double colon).
5) Let developer choose the separator by specifying dag property, but
default is still needed.
6) Any other option?

Please suggest which approach to take?


Regards,
-Tushar.

--001a114d7ec8910cda0525ce4ff6--