Live Coding on airdrag

legend:

• root-level: section-title
• indented-level: one of advice, note, comment
  • advice: yellow-StickIt + bottom-right
  • comment: red-cloud + top-right
  • note: blue-cloud + center

§

Contents

• 00:30: (1) IDE: install LiveShare plugin
  • 01:02: IDE: LiveShare not FOSS
  • 03:50: @@@ IDE: LiveShare installed
  • 04:05: act: navigate code with regexes
    • 04:32: …advice: regexes are nice
  • 04:53: …advice(IDE): vscode has ScreenCast mode
  • 05:58: act: Typing-hints for args
    • 06:10: …advice(code): fix styling & LoCs widths with black formatter
    • 06:36: …comment(IDE): "where am i? i need bookmarks in the code!"
• 07:13: (2) @@@ prepare styling & sort & fix (missing) imports
  • 07:34: @@@ …note(IDE): numba is missing from setup.cfg dependencies
  • 07:52: @@@ …numba+numpy advice
  • 08:49: @@@ IDE: google how to organize imports redproducibly (across IDEs)
    • 09:19: @@@ … comment(IDE): no googling failed! [were supposed to organize imports with a pipy library, eg isort, we better agree on some isort alternative as a team) NOTE: ensure isort works compatibly with black
  • 09:22: @@@ IDE: auto-formatting on save
    • 09:33: …advice(IDE): black+auto-save (with git-hooks as alternative)
    • 09:54: @@@ …comment(IDE): auto-organize imports will be added later (when??)
    • 10:01: @@@ …advice(IDE): do’nt place choice that modify the code in User-settings
    • 10:17: …advice(IDE): time spend on customizing your preferences is time spent wise
  • 11:20: Bookmarks plugin & IDEs commands
    • 11:51: …advice(IDE): VSCode + PyChar command pallet shortcuts
  • 13:57: commit(1): Style(black) & unused imports
    • 14:12: @@@ draft 1st commit msg
  • 15:03: IDE "Zen" mode (configs)
    • 15:15: …note(IDE): Zen mode shortcut
    • 15:30: Config VSCode Zen mode visible UI elements
    • 16:15: @@@ …comment(IDE) tooling is important
    • 16:38: …advice(IDE): VSCode + PyChar command pallet shortcuts (REPEATED)
• 16:50: (§3) Start Vectorizing outer loop
  • 18:20: commit(2): vectorize outer loop
• 18:58: (§4) Demeter Law for testable APIs
  • 19:21: note(git) …announce reshufling of vectorization commit after TC.
  • 24:15: @@@ …note(git): if future unit-test were to discover any bug…. the fix would be squashed with this commit.
  • 24:22: commit(3): Demeter’s Law
  • 24:48: advice(git): commit message widths
• 26:01: (§5) Start PyTesting
  • 26:38: setup.cfg testing dependencies
  • 27:05: @@@ fixture: dumy dataframe as input
  • 27:11: …advice(test): fixtures are good
  • 28:41: …comment(test): Perlin noise (not Sobol Sequences)
  • 29:04: @@@ Search for df columns
  • 31:14: CONNECTION BROKEN
  • 32:21: advice(test): unit-tests must not contain more data
  • 33:30: @@@ pytest dirs layout
    • 33:30: advice(test): google for pytest-layout+goodpractices
    • 34:02: @@@ delibarate on how to split unit-tests in test-files
    • 35:44: … note(test): test-files splitting not yet decided
  • 36:03: @@@ creating the unit-test function
  • 36:28: @@@ Yes! (reply to Thoma’s Q about how a fixture works)
  • 37:31: @@@ decide the number of dummy rows as a power of 2
  • 38:28: @@@ IDE: evaluate code interactively with [Shift+Enter]
  • 39:36: comment(build): explain scipy lib originates from my OS’s Python from
  • 40:00: Morph input data
    • 40:49: @@@ …note(IDE): most of the actions in this session heppen with the keyboard
    • 42:08: @@@ …note(IDE): plotting diagrams in the interactive panel
    • 43:06: comment(test): Sobols were an accident, it was Perlin Noise that i had in mind
  • 45:47: advice(code): temp variables aliases are ok
  • 46:13: Extract dummy df creation in a utility function (vs fixtures)
  • 48:02: note(IDE) …a CoPilot intervention
  • 49:17: @@@ advice(test): …configtest.py with this SO
  • 49:38: @@@ comment/advice(test): … apologies, utility functions need own files, (not configtest.py), meaning you need convert tests dir to a package with tests/__init__.py
  • 52:15: @@@ note(python): google: python+dunder+methods - but Python has no accessibility modifiers
  • 52:20: pytest configs
    • 53:06: @@@ advice(build): python projects have many ways to cofigure: config files: pyproject.toml, setup.cfg, setup.py(deprecated), other build libraries poetry, flit, hatch, build, …
    • 54:02: attempt for doctesting modules (and what is it?)
      • 54:09: @@@ comment(test): …spoiler: we will fail to doctest… due to many pre-existing errors
      • 54:43: @@@ advice(test): doctest mixes documentation with unit-testing, which if checked regularly, serves as an API call example code.
• 57:50: (§6) fix module import errors
  • 58:28: advice(test): a new Python interpeter facilitates debugging import errors
  • 1:00:27: @@@ advice(imports) 3 different import syntaxes:
    • import a_package.a_module as foo
    • from a_package import a_module [as foo]
    • from a_package.a_module import a_function [as bar]
  • 1:00:45: debugging & recreattting my problematic venv
  • 1:03:37: advice(IDE) mismatches between selected-interpeter vs terminal is the root of many problems
  • 1:03:55: @@@ Discovered tqdm unused package in imports
  • 1:04:22: @@@ git-knit(1) fixup unused-import fix into style-commit
    • 1:04:55: comment(git) big mistake mixing style & refactorings
  • 1:06:11: @@@ git-knit(2) reword style-commit msg about BUG
  • 1:06:24: @@@ advice(git) annotate commits that are known to contain a bug either by rewritting the commit msg or with git-notes
  • 1:07:48: relative imports
    • 1:08:31: advice(imports) import conventions for readability
    • 1:10:24: comment(imports) stuttered bc of a more succicnt import syntax
  • 1:10:42: commit(4) fix relative commits
    • 1:10:46: @@@ comment(git,IDE) Selecting changed code and with a shortcut…i staged the changes for the next commit
    • 1:11:20: comment(git) hounored sepration of commits this time…
    • 1:11:29: @@@ advice(git, IDE) Change your code feeely and before committing, preview changes and SELECTIVELY stage related ones in separate commits.
    • 1:12:30: @@@ note(git) notice how much time is spent on commit msg…that’s ok, it’s investment when programming in time
  • 1:12:40: Wandered studing rel-imports
    • 1:12:52: comment(imports) …waste a bit of time studying about rel-imports
  • 1:14:35: advice(code) keep side notes
  • 1:15:0: @@ git-knit(3): push package-fix earlier
    • 1:15:50: @@@ note(git) git rebase explanation
  • 1:16:20: @@@ commit(5) WIP PyTest
    • 1:16:26: @@@ advice(git) Instead of git-stash, a more verbose way is to commit a WIP-commit and git reset @~ )parent) later
    • 1:17:13: @@@ comment(imports) checking my modules once more with a pristine Python interpreter…
    • 1:17:27: comment(IDE) Exited my shared-terminal by habit….
    • 1:18:39: comment(IDE) Recreated my shared-terminal.
    • 1:19:13: git aliases for bash
    • 1:20:23: note: bookmarks & terminal misshups
    • 1:20:40: explain git-reset
• 1:22:11: (§7) Pytest part-2: seatbelt construction
  • 1:25:32: @@@ advice(code) always use pathlib for path manipulations
  • 1:26:38: Auto-organize imports on VSCode
  • 1:29:40: @@@ Recommence pytest
  • 1:30:07: code: pathlib manipulations
    • 1:29:40: advice(code) pathlib manipulations
  • 1:30:58: advice(data) accuracy-preserving file formats
  • 1:33:29: advice(data) union-typing supported by Python3.10
  • 1:35:42: advice(data) prefer defaults are false, OR stick to your convention
  • 1:41:00: Post-mortem debugging with pytest
  • 1:45:20: commit(6): seatbelt commit
• 1:47:15: (§8) @@ git-knit(2) fixup style changes & rewrite
  • 1:50:00: @@ git-knit(4) resolve conflicts while reshuffling
    • 1:50:45: Demeter Law, 2nd take
    • 1:52:02: advice(git) program in time needs disciplined commits
  • 1:53:03: reshuffle: preparing 1st conflict-resolved commit
    • 1:54:58: note(IDE) Ctrl+D to multi-select
    • 1:54:58: advice(git) tip: read the git-rebase prompt at the terminal to understand which changes needs to go
  • 1:59:57: reshuffle: preparing 2nd conflict-resolved commit
  • 2:02:56: Programming in time
• 2:05:27: (§9) @@ Chapter:Seatbelt finds vectorization BUG
  • 2:09:36: @@ git-knit(5): fixup timings into TC commit
  • 2:10:29: Study Vectorization BUG with git-diff
  • 2:12:33: LiveShare woes when git diffing
  • 2:15:00: Research vectorization bug (after resolving LiveShare woes)
  • 2:16:50: Draw internal loop
  • 2:18:40: Explain broadcasting for vectorization bug
  • 2:21:10: Resolved vectorization bug …"with many eyballs, all bug are shallow"
  • 2:23:28: Thomas asks for proof bug is solved
  • 2:24:48: x200 faster!
  • 2:25:40: …3k rows is NOT 1 hour but 10 hours/1Hz or 1H/10Hz
  • 2:26:30: Ask +10min to extend the live-coding session …but actually it took another +25min for the sesion to end!
  • 2:27.31: …insist on 1h—>36κ rows!!!
  • 2:29:20: …TC generates 2^15—32k rows, not 2
• 2:30:20: (§10) REFACT common code in temporary variables
  • 2:31:12: Auto-save enabled by user-prefs
  • 2:31:49: Continue REFACT extracting temp-vars (inner loop heights_ratio)
  • 2:33:10: Continue REFACT extracting temp-vars (v-veh in kmh)
  • 2:34:20: Continue REFACT extracting temp-vars (rel_wind)
  • 2:35:25: Continue REFACT linearity simplification
  • 2:36:48: …must leave a comment explaining this original formulas!!
  • 2:37:36: …f@ck!
  • 2:38:28: …Readability not Speed
  • 2:39:15: Continue REFACT radians
  • 2:40:20: …+BUG in Radians!!
  • 2:40:24: Continue REFACT reusing new series without indexing dataframes
  • 2:42:10: Radians-per-degrees saga!
    • 2:43:01: REALIZED there is a BUG
    • 2:45:51: RESOLVED the BUG it was the inverse radians-per-degrees
  • 2:46:41: …TCs must be instant, to TC on every keystroke
  • 2:47:15: Redo REFACTs new series in tmp var (lost when searching bug)
  • 2:47:50: FINISHED REFACT temp-vars, fast, readable & tested code.
  • 2:48:32: FINAL tmp-vars commit (same speed)
• 2:49:32: (§11) PUSH to pub repo (through an intermediate local)
  • 2:50:04: …mean repo
  • 2:51:39: END

TODO

• 11:16 mention where copilot appear in the time-in-video
• 11:20: compress bookmarks plugin installation
• 1:50:00, 2:05:25: repetition of "resolve-conflic" title

OpenShot Profile for MSTeams

description=MSTeams 16fps HD 1080p (1920x1080) frame_rate_num=16 frame_rate_den=1 width=1920 height=1080 progressive=1 sample_aspect_num=1 sample_aspect_den=1 display_aspect_num=16 display_aspect_den=9

This reshuffling situation explains perfectly well why we try so hard to keep git history "clean" and how this intermingles with testability & modularity (low coupling)