compound workflow with makefile style target

[BoPeng]

Running the following script

[analyze]
parameter: par=5
output: 'out.txt'

print(f'Analyze data with parameter {par}')
with open(_output, 'w') as out:
    out.write(f'Result from parameter {par}')

[summarize]
input: 'out.txt'    
print(f'Writing report with input out.txt')
report: input='out.txt', output='out.md'
    ## A summary report

[clean]
sh:
   echo Cleaning up
   rm -f out.txt out.md

with command

$ sos run test2 clean+summarize --par 20

generate this

INFO: Running clean:
INFO: Running summarize:
Writing report with input out.txt
Cleaning up
INFO: Workflow clean+summarize (ID=95ee5d7541554ad8) is executed successfully with 2 completed steps.

or

INFO: Running clean:
INFO: Running analyze:
Cleaning up
Analyze data with parameter 20
INFO: output:   out.txt
ERROR: [analyze (out.txt)]: Output target out.txt does not exist after the completion of step analyze
[clean+summarize]: 1 pending step: summarize

So basically the clean step is executed in parallel to the summarize workflow, and removes out.txt at an undetermined time.

Perhaps we should handle

clean+summarize

as

sos run clean
sos run summarize

instead of a single workflow with two parts.

[BoPeng]

The minimal test case is

[A]

[B]
input: None

and command

sos run test A+B

What is happening here is that step B is appended after A similar to a numerically indexed workflow, and then executed in parallel because of the input: None statement.