update instructions (#370)

.github/workflows/openhands-resolver.yml

@@ -8,7 +8,7 @@ on:
       max_iterations:
         required: false
         type: number
-        default: 50
+        default: 100
       macro:
         required: false
         type: string
@@ -195,6 +195,7 @@ jobs:
       - uses: r-lib/actions/setup-r-dependencies@v2
         with:
           extra-packages: any::roxygen2
+      - run: make install
 
       - name: Install OpenHands
         env:

.openhands_instructions

@@ -1,5 +1,7 @@
 # R Code Style Guide and Development Instructions
 
+## Overall guidance
+
 Your dev environment is *fully setup*. You do not need to install any packages.  They are all installed for you.
 
 Write the *minimal* code neccicary to solve the problem:
@@ -8,23 +10,20 @@ Write the *minimal* code neccicary to solve the problem:
 - Branching logic is bad
 - Convoluted error handling or edge case handling is bad
 - Use the fewest lines of code possible to solve the problem
-
-## Code Style
-
-- Use base R and core packages rather than tidyverse
-- Prefer data.table over data.frame for data manipulation
-- Use :: operator for package functions (e.g., data.table::data.table).
-  - Do not use `library` or `require`.  Import packages with ::
-- Avoid for loops - use vectorized operations
-- Use vapply instead of sapply
-- Keep functions small and focused
-- Write self-documenting code, minimize comments
-- Follow strict R code style guidelines
-- Write clean, lint-free, readable code
-- Do not use error handling
-- Do not use tryCatch / on.exit / try / withCallingHandlers / etc.
-- Do not write excessive if/then/else statements
-- Don't install or use new packages. You should already have everything you need.
+- Use the make commands in the Development to make sure your code works
+  - If these commands do not pass, you are not done with your work
+- Once lint/tests etc pass and you feel your work is complete, update [NEWS.md](NEWS.md) with a short description of your changes for the release log.
+- Before you do any work, make a plan!
+  - Your plan should be detailed, feel free to use a lot of tokens
+  - Use <planning> tags for your plan
+  - After you make a plan, score it with <plan-score> tags
+    - 0-100 on the likelihood that it fuffils the request
+    - 0-100 on complexity
+  - If the plan does not score above 80 on fulfills request, make a new plan
+  - If the plan does not score above 50 on complexity, make a new plan
+  - Repeat <planning> and <plan-score> until you have a good plan
+  - Finally, use a <reflection> tag to review your plan
+  - Once you've got a <reflection> tag, you are ready to start working on the request.
 
 ## Development Workflow
 
@@ -46,6 +45,38 @@ You shouldn't need to run the other make commands. If you update the README.rmd
 
 Do not run `make install`, `make update-test-fixtures`, `make release`, `make check-rev-dep`, `make check-win`, `make check-rhub`, `make check-many-preds`, `make check`,  `make preview-site`, `make dev-guide`.  If needed a human will run these commands.
 
+## Code Style
+
+- Use base R and core packages rather than tidyverse
+- Prefer data.table over data.frame for data manipulation
+- Use :: operator for package functions (e.g., data.table::data.table).
+  - Do not use `library` or `require`.  Import packages with ::
+- Avoid for loops - use vectorized operations
+- Use vapply instead of sapply
+- Keep functions small and focused
+- Write self-documenting code, minimize comments
+- Follow strict R code style guidelines
+- Write clean, lint-free, readable code
+- Do not use error handling
+- Do not use tryCatch / on.exit / try / withCallingHandlers / etc.
+- Do not write excessive if/then/else statements
+- Don't install or use new packages. You should already have everything you need.
+- Write clean, modular code
+- Use Roxygen2 to document your code
+- Write at least one unit test for your work
+- Don't add new packages. If you need to install an R package, something is wrong.
+- Prefer S3 classes over S4 and R5 classes when creating new classes
+- Avoid tidyverse packages unless absolutely necessary
+- ggplot2 and reshape2 are acceptable exceptions
+- Always use ggplot2 for plotting
+- Always use data.table::data.table for data manipulation
+- Avoid base data.frame and plyr completely
+- Prefer glmnet or xgboost for modeling tasks
+- Avoid for loops and sapply loops where possible
+- Look for built-in vectorized functions before writing loops
+- Consider vectorized operations from approved packages only if built-in functions aren't available
+- make sure all files end in newlines
+
 ## Package Structure
 
 - Put R code in R/
@@ -62,7 +93,6 @@ Do not run `make install`, `make update-test-fixtures`, `make release`, `make ch
 - Be sure to test error conditions, especially cases that should throw errors
 - Use testthat framework
 
-
 ## Getting Help
 
 These file contain information that will help you understand the code structure and workflow:
@@ -71,3 +101,9 @@ These file contain information that will help you understand the code structure
 - [Makefile](Makefile) - The Makefile. Contains many useful commands, but heed the instructions in this file: you won't need to use most of those make commands.
 - [.openhands_instructions](.openhands_instructions) - This file. Contains instructions for you as a developer.
 - [project-tree.txt](project-tree.txt) - A nice clean package directory, ignoring files you dont need to edit. Use this file to navigate the package and understand which files you need to edit. Feel free to re-make this file with `make project-tree.txt` if you've made file or directory changes to see what's changed.
+- Feel free to use `git` commands to see what's changed recently. Don't forget you are non-interactive so you need to run non-interactive commands like `git status` or `git diff` with `-c` or `--no-pager` flags.  E.g.:
+  - `git log --no-merges --pretty=format:"%H|%an|%ad|%s" --date=iso -n 10` to see recent commits
+  - `git rev-parse HEAD` to see the current commit hash
+  - `git diff-tree --no-commit-id --name-only -r HEAD` to see the files that have changed in the last commit
+  - `git log -1 --pretty=%B` to see the commit message of the last commit
+  - `git diff --name-only HEAD~1 HEAD` List files changed between two commits