Transcript based testing¶
A transcript is both the input and output of a successful session of a
cmd2-based app which is saved to a text file. With no extra work on your
part, your app can play back these transcripts as a unit test. Transcripts can
contain regular expressions, which provide the flexibility to match responses
from commands that produce dynamic or variable output.
Creating a transcript¶
A transcript can automatically generated based upon commands previously executed in the history:
(Cmd) help ... (Cmd) help history ... (Cmd) history 1:2 -t transcript.txt 2 commands and outputs saved to transcript file 'transcript.txt'
This is by far the easiest way to generate a transcript.
Make sure you use the poutput() method in your
cmd2 application for generating command output. This method
cmd2.Cmd class ensure that output is properly redirected when redirecting to a file, piping to a shell
command, and when generating a transcript.
Here’s a transcript created from
(Cmd) say -r 3 Goodnight, Gracie Goodnight, Gracie Goodnight, Gracie Goodnight, Gracie (Cmd) mumble maybe we could go to lunch like maybe we ... could go to hmmm lunch (Cmd) mumble maybe we could go to lunch well maybe we could like go to er lunch right?
This transcript has three commands: they are on the lines that begin with the prompt. The first command looks like this:
(Cmd) say -r 3 Goodnight, Gracie
Following each command is the output generated by that command.
The transcript ignores all lines in the file until it reaches the first line that begins with the prompt. You can take advantage of this by using the first lines of the transcript as comments:
# Lines at the beginning of the transcript that do not ; start with the prompt i.e. '(Cmd) ' are ignored. /* You can use them for comments. */ All six of these lines before the first prompt are treated as comments. (Cmd) say -r 3 Goodnight, Gracie Goodnight, Gracie Goodnight, Gracie Goodnight, Gracie (Cmd) mumble maybe we could go to lunch like maybe we ... could go to hmmm lunch (Cmd) mumble maybe we could go to lunch maybe we could like go to er lunch right?
In this example I’ve used several different commenting styles, and even bare text. It doesn’t matter what you put on those beginning lines. Everything before:
(Cmd) say -r 3 Goodnight, Gracie
will be ignored.
If we used the above transcript as-is, it would likely fail. As you can see,
mumble command doesn’t always return the same thing: it inserts random
words into the input.
Regular expressions can be included in the response portion of a transcript, and are surrounded by slashes:
(Cmd) mumble maybe we could go to lunch /.*\bmaybe\b.*\bcould\b.*\blunch\b.*/ (Cmd) mumble maybe we could go to lunch /.*\bmaybe\b.*\bcould\b.*\blunch\b.*/
Without creating a tutorial on regular expressions, this one matches anything
that has the words
lunch in that order. It doesn’t
to appear in the output, but it does work if
mumble happens to add words to the beginning or the end of the output.
Since the output could be multiple lines long,
cmd2 uses multiline regular
expression matching, and also uses the
DOTALL flag. These two flags subtly
change the behavior of commonly used special characters like
$, so you may want to double check the Python regular expression
If your output has slashes in it, you will need to escape those slashes so the stuff between them is not interpred as a regular expression. In this transcript:
(Cmd) say cd /usr/local/lib/python3.6/site-packages /usr/local/lib/python3.6/site-packages
the output contains slashes. The text between the first slash and the second slash, will be interpreted as a regular expression, and those two slashes will not be included in the comparison. When replayed, this transcript would therefore fail. To fix it, we could either write a regular expression to match the path instead of specifying it verbatim, or we can escape the slashes:
(Cmd) say cd /usr/local/lib/python3.6/site-packages \/usr\/local\/lib\/python3.6\/site-packages
Be aware of trailing spaces and newlines. Your commands might output trailing spaces which are impossible to see. Instead of leaving them invisible, you can add a regular expression to match them, so that you can see where they are when you look at the transcript:
(Cmd) set prompt prompt: (Cmd)/ /
Some terminal emulators strip trailing space when you copy text from them. This could make the actual data generated by your app different than the text you pasted into the transcript, and it might not be readily obvious why the transcript is not passing. Consider using Output redirection to the clipboard or to a file to ensure you accurately capture the output of your command.
If you aren’t using regular expressions, make sure the newlines at the end of your transcript exactly match the output of your commands. A common cause of a failing transcript is an extra or missing newline.
If you are using regular expressions, be aware that depending on how you
write your regex, the newlines after the regex may or may not matter.
\Z matches after the newline at the end of the string, whereas
$ matches the end of the string or just before a newline.
Running a transcript¶
Once you have created a transcript, it’s easy to have your application play it
back and check the output. From within the
$ python example.py --test transcript_regex.txt . ---------------------------------------------------------------------- Ran 1 test in 0.013s OK
The output will look familiar if you use
unittest, because that’s exactly
what happens. Each command in the transcript is run, and we
output matches the expected result from the transcript.
If you have set
allow_cli_args to False in order to disable parsing of
command line arguments at invocation, then the use of
to run transcript testing is automatically disabled. In this case, you can
alternatively provide a value for the optional
constructing the instance of your
cmd2.Cmd derived class in order to
cause a transcript test to run:
from cmd2 import Cmd class App(Cmd): # customized attributes and methods here if __name__ == '__main__': app = App(transcript_files=['exampleSession.txt']) app.cmdloop()