Strategy for taking in arguments when app config is not an option

[ecurtin]

A lot of blood, sweat, and tears went into ensuring that the Cassandra configuration process could be separated from source control so that you didn't have to check in a bunch of private data about your database connection: ips, passwords, SSL certs, etc.

However, a large portion of potential users (i.e. Bluemix) are in an environment where they don't even have access to administrate the instance, so using configuration objects won't fly.

For this use case, it would be helpful to have an alternative way to configure objects so that these types of users can still configure their operator without bugging an admin. Some options:

• Good ol' fashioned SPL arguments
• Parameter file that gets baked in at compile time
• Some fanciness the Streams team was talking about that'll prompt for user/pass at submit time.

[natashadsilva]

The fanciness I was referring to is not really a separate option, you would still be using parameters to the operator. It is a best practice of sorts that allows you to have parameters that are not specified till the application is ready to be run.
You are trying to avoid this:

() as Sink = CassandraSink(Tuples){
 param
user: "user1";
password: "password";
}

By using something called a submission time value. It looks like this:

() as Sink = CassandraSink(Tuples){
 param
user: getSubmissionTimeValue("cassandra_username");
password: getSubmissionTimeValue("password");
}

So you recommend to users that they make the actual values for username and password (or anything else) a parameter to the final Streams application. They do this using submission time values instead of leaving the username/password in the code. Whoever runs the application will now have to specify two parameters called "cassandra_user" and "password" when submitting the job. On Bluemix, they would be prompted for those parameters:

Submission time values are a really handy way of playing around with parameter values without having to recompile your application. You can just specify a different value every time.

This could be expanded such that you could reuse the application to connect to different hosts without having to recompile or change the application config:

() as Sink = CassandraSink(Tuples){
 param
user: getSubmissionTimeValue("cassandra_username");
password: getSubmissionTimeValue("password");
hostname: getSubmissionTimeValue("hostname", "default_host"); 
port: (int32)getSubmissionTimeValue("port", "9044"); //cast returned value to an int.
}

As you can see above, the second, optional parameter can be used to specify a default value.

[ecurtin]

This issue is partially sastified by providing means for JSON String config as SPL parameters in #7