## CodeRunner Documentation (V3.1.0)

When the template is a per-test template and a TemplateGrader is selected, the output from the program must be a JSON string that defines one row of the test results table. [Remember that per-test templates are expanded and run once for each test case.] The JSON object must contain at least a 'fraction' field, which is multiplied by TEST.mark to decide how many marks the test case is awarded. It should usually also contain a 'got' field, which is the value displayed in the 'Got' column of the results table. The other columns of the results table (testcode, stdin, expected) can, if desired, also be defined by the template grader and will then be used instead of the values from the test case. As an example, if the output of the program is the string

{"fraction":0.5, "got": "Half the answers were right!"}


half marks would be given for that particular test case and the 'Got' column would display the text "Half the answers were right!".

For even more flexibility the result_columns field in the question editing form can be used to customise the display of the test case in the result table. That field allows the author to define an arbitrary number of arbitrarily named result-table columns and to specify using printf style formatting how the attributes of the grading output object should be formatted into those columns. For more details see the section on result-table customisation.

Writing a grading template that executes the student's code is, however, rather difficult as the generated program needs to be robust against errors in the submitted code. The template-grader should always return a JSON object and should not generate any stderr output.

It is recommended that template graders be written in Python, regardless of the language in which the student answer is written, and that Python's subprocess module be used to execute the student code plus whatever test code is required. This ensures that errors in the syntax or runtime errors in the student code do not break the template program itself, allowing it to output a JSON answer regardless. Some examples of per-test template graders are given in the section Template grader examples.

Sometimes the author of a template grader wishes to abort the testing of the program after a test case, usually the first, e.g. when pre-checks on the acceptability of a student submission fail. This can be achieved by defining in the output JSON object an extra attribute abort, giving it the boolean value true. If such an attribute is defined, any supplied fraction value will be ignored, the test case will be marked wrong (equivalent to fraction = 0) and all further test cases will be skipped. For example:

{"fraction":0.0, "got":"Invalid submission!", "abort":true}

Note to Python programmers: the Python boolean literals are True and False but the JSON boolean literals are true and false. The superficial similarity of JSON objects to Python dictionaries is also confusing. Usually, in Python, you will generate the JSON objects using the dumps function in the json module, passing it a Python dictionary, further adding to the confusion!

For example, to generate the above JSON output (with lower case t in true) you would write

json.dumps({"fraction":0.0, "got":"Invalid submission!", "abort":True})