Are Rewrites Always a Bad Idea?

Transcript

1. Are Rewrites always a Bad Idea? Adele Carpenter @iam_carpenter AR E R EWRITES ALWA YS A BAD IDEA?
2. This project sucks Diversity People Vectors by Vecteezy
3. Small steps is the way forward! Always? Diversity People Vectors by Vecteezy
4. PROJECT PRODUCT Custom Custom Custom Custom
5. Item selection service Assessment Delivery Reporting service Data export service C++ Implementation of Item Response Theory
6. Item Response Theory (IRT) AR E R EWRITES ALWA YS A BAD IDEA? Least Difficult Items Most Difficult Items
7. Item Response Theory (IRT) AR E R EWRITES ALWA YS A BAD IDEA? Least Difficult Items Most Difficult Items Assessment
8. Assessment
9. Student A - lower ability Student C - higher ability Student B - moderate ability
10. Assessment Delivery Reporting service Data export service - Hard to verify changes - No logging - Non human-readable requests
11. aPar bPar CurSeg Theta WarmTheta Clu Ctg Naming is Hard AR E R EWRITES ALWA YS A BAD IDEA?
12. - Black box - Hard to reason about and debug - Trivial changes take a long time Problems with the Current Service AR E R EWRITES ALWA YS A BAD IDEA?
13. - Stateless - Well-defined scope or footprint - Small part of the entire project What makes it a Good Rewrite Candidate? AR E R EWRITES ALWA YS A BAD IDEA?
14. AR E R EWRITES ALWA YS A BAD IDEA? Photo by Chris Lawton on Unsplash - There is a desire to functionally change the application 0. Reasons to Rewrite
16. AR E R EWRITES ALWA YS A BAD IDEA? Photo by Chris Lawton on Unsplash - There is a desire to functionally change the application - There are no small steps to be found 0. 1. Reasons to Rewrite
17. Photo by Sammie Chaffin on Unsplash
18. Question Yourself - Bias FOR the rewrite (most common!!): - Why should I keep the existing code base? - What opportunities exist to improve it? AR E R EWRITES ALWA YS A BAD IDEA? Photo by Chris Lawton on Unsplash - Bias AGAINST: - Are the improvement opportunities really there? - How much work are these “improvements” compared to a rewrite? Photo by Ilkka Kärkkäinen on Unsplash
20. AR E R EWRITES ALWA YS A BAD IDEA? Photo by Chris Lawton on Unsplash - There is a desire to functionally change the application - There are no small steps to be found - The skills are not in the team and you don’t wish to acquire them 0. 1. 2. Reasons to Rewrite
21. No wish to acquire the required skills AR E R EWRITES ALWA YS A BAD IDEA? Photo by Divani on Unsplash Photo by an_vision on Unsplash Photo by Meritt Thomas on Unsplash
23. AR E R EWRITES ALWA YS A BAD IDEA? Photo by Chris Lawton on Unsplash - There is a desire to functionally change the application - There are no small steps to be found - The skills are not in the team and you don’t wish to acquire them - The service is small and scope well- defined 0. 1. 2. 3. Reasons to Rewrite
24. Fig tree in the forest by Tatters, used under CC BY 2.0 license https://www.flickr.com/photos/tgerus/24057488061
25. Load Balancer Starting position Early days Halfway Nearly finished Finished Internet icons created by Freepik - Flaticon Load Balancer Load Balancer Load Balancer Load Balancer
26. AR E R EWRITES ALWA YS A BAD IDEA? Strangler Fig Application Go down the rabbit hole from Martin Fowler’s original blog
27. Spike - Experiment to explore a solution to a new problem - “End-to-end but very thin” - Best bet is a functional spike - Can help you figure out how to deliver value early AR E R EWRITES ALWA YS A BAD IDEA? Photo by Chris Lawton on Unsplash Photo by Ayoola Salako on Unsplash
28. Reasons to Rewrite - There is a desire to functionally change the application - There are no small steps to be found - The skills are not in the team and you don’t wish to acquire them - The service is small and scope well- defined AR E R EWRITES ALWA YS A BAD IDEA? Photo by Chris Lawton on Unsplash 0. 1. 2. 3.
29. Item information calculation Student ability estimation (Online) Item calibration Content balancing Item selection IN scope? YES YES YES NO NO AR E R EWRITES ALWA YS A BAD IDEA? Feature Audit Item selection service
30. You are not special - If you think you are, you’re at peak Dunning-Kruger #sorrynotsorry - It’s probably “so complicated” for a reason - Treat the first phase as a discovery or feasibility stage - Prepare yourself to walk away AR E R EWRITES ALWA YS A BAD IDEA? Photo by Chris Lawton on Unsplash Photo by Hobi industri on Unsplash
31. Item selection service Item information calculation Student ability estimation (Online) Item calibration Content balancing Item selection IN scope? YES YES YES NO NO AR E R EWRITES ALWA YS A BAD IDEA? Plan for version 1.0
32. Development Begins AR E R EWRITES ALWA YS A BAD IDEA? Photo by Divani on Unsplash - Kotlin / Spring Boot
33. Choosing the Tech Stack AR E R EWRITES ALWA YS A BAD IDEA? Photo by Divani on Unsplash - MultiK could work - But we were only dealing with 2D arrays - Kotlin / Spring Boot - Numerical packages in Kotlin - MultiK and SMILE - SMILE overpowered
34. data class WaffleSales(val type: String, val price: Int, val qtySat: Int, val qtySun: Int) val stroop = WaffleSales("stroop", price = 2, qtySat = 15, qtySun = 20) val belgian = WaffleSales("belgian", price = 3, qtySat = 25, qtySun = 18) val weekendSalesData = listOf(stroop, belgian)
35. data class Saturday(val qtySold: Int) data class Sunday(val qtySold: Int) data class WaffleSales(val type: String, val price: Int, val sat: Saturday, val sun: Sunday) val stroop = WaffleSales("stroop", price = 2, Saturday(15), Sunday(20)) val belgian = WaffleSales("belgian", price = 3, Saturday(25), Sunday(18)) val weekendSalesData = listOf(stroop, belgian)
36. data class Saturday(val qtySold: Int) data class Sunday(val qtySold: Int) data class WaffleSales(val type: String, val price: Int, val sat: Saturday, val sun: Sunday) val stroop = WaffleSales("stroop", price = 2, Saturday(15), Sunday(20)) val belgian = WaffleSales("belgian", price = 3, Saturday(25), Sunday(18)) val weekendSalesData = listOf(stroop, belgian) val dailySales = listOf( weekendSales.sumOf { it.price * it.sat.qtySold }, weekendSales.sumOf { it.price * it.sun.qtySold } ) println(dailySales) // [105, 94]
37. e
38. e = 1 + 1 1 + 1 2 + 1 2 × 3 + 1 2 × 3 × 4 + … + 1 n! e ≈ 2.71828182845904523536 0287471352662497757247093 6999595749669676277240766 3035354759457138217852516 6427427466391932003059921 8174135966290435729003342 95260… d dx ex = ex
39. val eToPower = exp(3.0) // ~ 2.718 ^ 3 println(eToPower) // 20.0855 val naturalLog = ln(eToPower) // Log to base e of 20.08 println(naturalLog) // 2.9997 (about 3) // other handy functions // pow() abs() sqrt()
40. Photo by Jamie Street on Unsplash
41. Photo by Jamie Street on Unsplash
43. Are Rewrites Always a Bad Idea? AR E R EWRITES ALWA YS A BAD IDEA?
44. *But they can be a good idea sometimes. YES!* Are Rewrites Generally a Bad Idea? AR E R EWRITES ALWA YS A BAD IDEA?
45. When are they not? - There is a desire to functionally change the application - There are no small steps to be found - The skills are not in the team and you don’t wish to acquire them - The service is small and scope well- defined AR E R EWRITES ALWA YS A BAD IDEA? Photo by Chris Lawton on Unsplash 0. 1. 2. 3.
46. - Tech stack: made a good enough choice in the context of your company and team - Unfamiliar technology = more risk - Add value as early as possible - Conduct experiments - What so your users / stakeholders think? If you really must rewrite AR E R EWRITES ALWA YS A BAD IDEA?
47. Thank you! Adele Carpenter @iam_carpenter AR E R EWRITES ALWA YS A BAD IDEA?