JMeter: Which to use: CSV, Array, or Array List?

Question

First let me preface this with 1) I'm not a developer by nature, and 2) I've only been using JMeter for less than 1 month.

I am performing basic CRUD testing of our APIs to/from our database. Due to a table constraint, each record in the table under test must have a unique typeId and groupId where the typeId/groupId combination is pre-defined elsewhere.

I need to perform the following upon a basic Create:

Path: /.../.../.../crud/QualificationExpression Method: POST

{
    "id": null,
    "displayName": "crudTest: ${__Random(1,10000000, randomNumber)}",
    "functionName": null,
    "displayableSourceExpression": "TRUE",
    "clobObjId": null,
    "typeId": xxx,
    "groupId": yyy,
    "typeCode": null
}

I'm trying to figure out what is the best way to do this:

CSV file
Array
Array List

Searching the forums and other online resources, when people use a CSV they're typically using it to store username|passwords for random logins. This results in them using Parameters instead of a Post Body.

The Array examples all seem to contain a single column array where they pull a random value from within the array. However, I have a pair: if typeId = x, then objectId must = y.

Then I see an Array List sample where I'm totally confused:

ArrayList listOfPairs = new ArrayList();
    ArrayList pair1 = new ArrayList();
       pair1.add(203);
       pair1.add(303);
       listOfPairs.add(pair1);
    ArrayList pair2 = new ArrayList();
       pair2.add(204);
       pair2.add(304);
       listOfPairs.add(pair2);
    ArrayList pair3 = new ArrayList();
       pair3.add(205);
       pair3.add(305);
       listOfPairs.add(pair3);

    for (int i = 0; i < listOfPairs.size(); i++) {
        ArrayList pair = (ArrayList)listOfPairs.get(i);
        System.out.println("pair:" + pair.get(0) + " " + pair.get(1));
    }

My desire would be to not have a CSV file and it would be best to have the test plan .jmx be stand-alone without relying on "payload" files being present.

I'm hoping those with more experience and knowledge could provide some guidance and insight.

Dmitri T Dmitri T · Accepted Answer · 2016-03-10T05:47:15

Go for Functions, from performance perspective it's better to use built-in components where possible.

If you:

need to have pre-defined (not random) typeId and groupId combination - you need to keep these pairs somewhere
don't want external files - you need to store these files in Test Plan itself, i.e. in User Defined Variables
need typeId and groupId pairs to be unique - the most obvious solution is using current virtual user number as prefix or postfix like:
- in User Defined Variables:
- in Request Body use __threadNum() and __V() functions combination like:
```
"typeId": "${__V(typeId_${__threadNum})}",
"groupId": "${__V(groupId_${__threadNum})}",
```
plan to run test for several iterations - use __counter() function as an extra postfix

See How to Use JMeter Functions articles series for comprehensive information on above and other JMeter functions.

JMeter: Which to use: CSV, Array, or Array List?

2 Answers