Measure Before Optimizing

Use timings/query plans/profiles to locate the bottleneck before changing code.

Learning contract

What you must be able to do

Objectives

Explain Measure Before Optimizing using a concrete runtime/data-flow model.Implement the smallest working example and predict its behavior before execution.Diagnose a controlled failure related to Measure Before Optimizing using evidence rather than random edits.Transfer the concept into the TaskFlow project without copying the finished example.

Mental model

Understand the idea before syntax

Turn production quality into observable, testable engineering rather than a checklist.

D9.7 — Evidence loop

predict

implement

run

observe

locate first mismatch

repair

re-run

When the result differs from your prediction, debugging starts at the first boundary where evidence diverges—not at the last visible symptom.

Before you code

Establish a healthy baseline

01
Open a scratch file or the relevant TaskFlow package.
02
Run the existing project/tests once so you know the baseline is healthy.
03
State what evidence will prove success before changing code.

Instructor code-along

Small change, immediate evidence

01
Write the goal in observable terms before coding: by the end of this lesson you should be able to demonstrate “Measure Before Optimizing” in a running TaskFlow-oriented example.
02
Read the example once without running it. Predict the exact output, state change, request, or compiler behavior and write that prediction down.
03
Type the example yourself rather than pasting it. Run the narrowest command that exercises only this concept, then compare observed evidence with your prediction.
04
Change one assumption at a time. Keep the change small enough that you can explain why the result changed.
05
Perform the intentional failure lab: Add caching first, hide the slow query, then discuss the invalidation complexity you introduced.
06
Read the resulting compiler error, stack trace, HTTP response, browser evidence, database state, or CI log from the first line that is actually relevant to your code.
07
Repair only the root cause. Re-run the same verification so you can prove the repair changed the observed evidence.
08
Close the reference and rebuild the important part from memory. If you cannot explain why each boundary exists, revisit the smallest missing concept rather than memorizing the final code.

textWorking example

baseline GET /tasks: 1.8s
EXPLAIN ANALYZE → seq scan
add targeted index
retest: 140ms

Expected evidence

The optimization is justified by before/after evidence.

Break it on purpose

Add caching first, hide the slow query, then discuss the invalidation complexity you introduced.

Do not fix it immediately. Capture the compiler error, request/response, stack trace, database state, browser output, or CI log first. Debugging skill grows from comparing evidence with a prior prediction.

Investigation checklist

What did you predict?
What actually happened?
At what boundary do those first differ?
What is the smallest piece of evidence that proves the failure lives there?
Can you fix the cause without changing unrelated layers?
Can you add a regression test or repeatable verification?

Transfer challenge

Now work without the transcript

Apply Measure Before Optimizing to one new TaskFlow requirement. Do not reopen the example until you have a failing attempt, an explanation of the failure, and a specific question you cannot answer.

Exit ticket

Continue only when these are true

Mastery evidence

I can explain why Measure Before Optimizing exists.I can predict the example before running it.I can reproduce the core behavior from memory.I can create and diagnose the intentional failure.I can name what would change in a production system.

Lesson progress

Record this deep-dive lesson

Mark complete only after the transfer challenge and exit ticket.