Videos / Web Dev

Episode 10 of 16 · Web development, end to end

Turborepo and the monorepo: one repository, many packages

Chapters

1. Two copies of one type0:00
2. The publish delay0:20
3. One pull request, not two0:37
4. And a graph0:55
5. A monorepo is not a monolith1:09
6. So what is turbo1:30
7. The shelf layout1:45
8. One field does the work2:02
9. Linked, not downloaded2:19
10. The protocol and the lockfile2:36
11. Compiled, or just in time2:54
12. Resolution done, order open3:12
13. You cannot ice it first3:31
14. What a task is3:47
15. One caret does it4:03
16. Walking the graph4:21
17. And what is independent4:40
18. The empty dependsOn4:56
19. The fingerprint5:12
20. What goes in the hash5:28
21. Replaying a build5:47
22. Declaring the outputs6:04
23. The input list that lies6:22
24. Two rules for the cache6:41
25. The empty disk7:01
26. The shared store7:18
27. Only the changed packages7:36
28. Not starting the job at all7:55
29. A cache is a trust boundary8:14
30. Three mechanisms8:33
31. One front door8:49
32. Writing it down9:05
33. Blast radius9:23
34. The import that should not9:41
35. What one repo did not fix10:00
36. What it charges10:19
37. The whole thing10:38

Turborepo and the monorepo: one repository, many packages. Two repositories. One shared type definition, copied into both. Someone edits it in one place, and the other keeps compiling happily against the old shape until production disagree

Mark as watchedSaved in this browser only.