Practical Workflow Automation
Build a job that runs without you watching it. Map a workflow before automating it, keep every change safe to re-run, organise files and process everyday records, work with a service that is sometimes unavailable, and notice when the job stops running at all. Assumes basic Python functions, loops, and dictionaries. Approximately 8-12 hours including practice, an unvalidated planning estimate rather than a measured completion time.
- Modules
- 6
- Lessons
- 18
- Estimated time
- About 8 hours of lessons
- Before you start
- Assumes basic Python: functions, loops and dictionaries.
Times are planning estimates from each lesson's stated length, not measured completion times.
Reading and practice are open to everyone. Saving your progress needs an account with access to this course.
What the practice looks like
- In A batch of records, plus what the previous run finished
- Each record
- Finished by the previous run: Skip it
- Handled successfully: Add it to processed
- Rejected for good (RecordRejected): Quarantine it and keep going
- Service unavailable or timed out: Retry, at most 2 attempts in all, then stop the run as "partial"
- Out A run record: status, processed, skipped, quarantined and reason
Course roadmap
Modules are in the recommended order, and every lesson is open from day one.
Plan an Automation
Decide what is worth automating, write down what one run consumes and produces, and prove the plan with a dry run before anything changes.
Organize Files
Read a folder without being fooled by it, plan renames that cannot collide, and tell a duplicate file from one that merely shares a name.
Process Everyday Data
Read a batch of records without trusting the shape it arrived in, let a bad record fall out of the run instead of stopping it, and turn what is left into a report whose numbers add back up to the count you started with.
Connect Services
Work with a service you cannot supervise: give every run a request budget and a deadline, test against responses you recorded rather than a network, and page through results without ever making an unbounded number of requests.
Make It Dependable
Write down what each run did in a form the next run can read, resume after a partial failure instead of starting again, and notice the job that stopped running at all.
Capstone: The Order Digest
Assemble everything into one scheduled reporting workflow, prove it survives being interrupted and re-run, and package it so somebody who has never seen it can run, schedule, and recover it.