Documentation on developing this library

AUTHOR.md

@@ -1 +1 @@
-This library was made by Ethan Uppal and Utku Melemetci.
+This library was made by Ethan Uppal, Jess Wang, and Utku Melemetci.

README.md

@@ -17,17 +17,14 @@ This repository hosts CEV's implementation of Iterative Closest Points (ICP) as
 Please see [LICENSE](LICENSE). Note that all code is licensed under the MIT license, with the exception of some additions to `doxygen-style.css`. Files with copyright headers denoting specific copyright holders belong to said holders. Files under `book/asset` are from [doxygen_theme_flat_design](https://github.com/kcwongjoe/doxygen_theme_flat_design) under its MIT license (again with the exception of some additions to `doxygen-style.css`). Files without copyright headers are under the MIT license, Copyright (c) 2024-2025 Cornell Electric Vehicles.
 
 ## Install
-
 You can view installation instructions at [INSTALL.md](INSTALL.md).
 
 ## Usage and Documentation
-
 We host the usage information and documentation at [cornellev.github.io/icp/](https://cornellev.github.io/icp/).
 Please see there for information on how to download and how to use or extend the library.
 
 You can build the documentation yourself locally with `make docs`.
 The main page will be located at `docs/index.html` relative to the project root.
 
 ## Versions
-
 Version information can be found in the [releases](https://github.com/cornellev/icp/releases) page.

book/developing.md

@@ -0,0 +1,70 @@
+# Developing
+
+This document explains how to effectively develop `icp`. It covers development environment setup, the general repository structure, how to add new ICP methods, and how to add benchmarks and tests.
+
+## Development environment
+
+### Dependencies
+`libicp` itself can be built and installed with only Eigen as a dependency---this is by design. However, to effectively develop the project, you'll need a few extra things.
+
+1. [libcmdapp2](https://github.com/cornellev/libcmdapp2), a command-line parsing library (built by Ethan) which we use for the 2D visualization.
+2. [sdl-wrapper](https://github.com/cornellev/sdl-wrapper), a C++ wrapper for SDL to make things easier for the visualization.
+3. [simple-test](https://github.com/cornellev/simple-test), a header-only testing library.
+4. [PCL](https://pointclouds.org/) 1.12.1 (ideally), a point cloud processing library. We actually only use this to parse 3D point clouds for testing.
+
+Each of the links above should have installation instructions. If you have everything installed correctly, you should be able to [build](#building) the project and not have any errors.
+
+### Tooling setup
+I'll mainly cover VSCode here as it's relatively popular, but other IDEs shouldn't be too difficult.
+
+You'll need the [clangd](https://marketplace.visualstudio.com/items?itemName=llvm-vs-code-extensions.vscode-clangd) and [clang-format](https://marketplace.visualstudio.com/items?itemName=xaver.clang-format) extensions. You may also need to install `clang-format` itself. Please install at least version 19 as some of the rules depend on recent versions.
+
+The build system, by default, will generate the `compile_commands.json` file that `cland` relies on. However, it will be in `build/`. To configure VSCode to pick up on this file, you can add the following to your `.vscode/settings.json`
+
+```json
+{
+    "clangd.arguments": [
+        "-compile-commands-dir=build"
+    ],
+}
+```
+
+## Repository structure
+```
+.
+├── bench - code for benchmarking
+├── book - documentation and files for the website
+├── common - common code across benches, tests, library, and vis
+├── ex_data
+│   ├── ply - .ply example data (3D)
+│   └── scan{n} - .csv example data (2D)
+├── include - library header files
+├── lib - library source files
+├── script - various utility scripts
+├── tests - unit test code
+└── vis - 2D visualization app
+```
+
+A few things to note:
+1. `tests` also contains `test_ply.cpp`. This isn't really a test, but a program you can run to look at the output from aligning two `.ply` scans. We might want to mvoe this in the future.
+2. A lot of the library code is templated. So you may need to look in `include` for some of the core functionality.
+3. You'll probably be doing most of your work in `include` and `lib`. These are the core folders.
+
+## Building
+`CMakeLists.txt` is the main build file. However, for convenience, there is a `Makefile` that shortens the `cmake` commands. Just typing `make` will build everything: the library, visualization, benchmarks, and tests.
+
+There are some extra options you should be aware of. Setting `OPT=[Release | Debug]` (i.e. `make OPT=Release`) changes the optimization level. `USE_SANTIZERS=ON` will enable the address and undefined behavior sanitizers, which can be useful when debugging UB. `CI=ON` will run a CI build, which compiles with warnings as errors and enables the sanitizers.
+
+Make sure to check the `Makefile` for other utilities you might find useful. You may be interested in `make tidy` and `make view`.
+
+## Testing
+You can run `make test` to run the 2D tests and `make test_3d` to run the 3D tests. Ideally, in the future, we combine these into one executable, but we haven't done this yet.
+
+### Adding Tests
+To add tests, you can follow the examples already in `test.cpp` and `test3d.cpp`. If you want to create a new test file, it's probably best that you refactor and bring all the tests under one executable.
+
+## Benchmarking
+You can use `make bench` to run the benchmarks. They'll run 2D ICP on the example data. There are currently no 3D ICP benchmarks.
+
+## Continuous Integration
+The CI system is defined in `.github/workflows/ci.yaml`. It's pretty standard: it just runs the tests on MacOS and Linux.

book/writing_icp.md

@@ -1,104 +1,90 @@
-<!--  
-Copyright (C) 2025 Ethan Uppal.
-SPDX-License-Identifier: MIT
---->
-\page write_icp_instance Writing an ICP Instance
+# Contributing: Adding a New ICP Instance
 
-\tableofcontents
+This section explains how to implement and document a new **ICP instance**.
 
-To write an ICP instance, create a C++ source file with at least the following:
+---
 
-1. A `final` class that inherits from `public icp::ICP`
-2. A static initialization variable (described below)
+## 1. Overview
 
-It is highly recommended you also provide documentation for your ICP instance in the format described here.
+An ICP instance is a specific implementation of the Iterative Closest Point algorithm. Each instance should be:
 
-\section core_func_sec Core Functionality
+- A `final` C++ class that inherits from `icp::ICP`
+- Properly registered so users can instantiate it
 
-The class must define the following behavior:
+---
 
-- A constructor that calls the `icp::ICP` constructor
-- An overridden destructor (which may do nothing)
-- `void iterate() override`
+## 2. File and Class Requirements
 
-In `iterate`, the point clouds are given by the instance variables `a` and `b`.
-There is also the `match` instance variable, allocated to have size `a.size()`, which cannot be assumed to contain any definite values.
-At the end of `iterate`, the `transform` instance variable should have been updated (although the update may be zero).
+Create a new `method_xd.cpp` file in `lib/icp/impl/`, and implement your instance class.
 
-Optionally, the class can override:
+### Required
 
-- `void setup() override`
+- Class must be `final` and inherit from `public icp::ICP`
+- Must implement `void iterate() override`
+- Must provide a constructor that calls the `icp::ICP` constructor
+- A destructor (can be empty)
 
-`setup` is invoked upon the user call to `ICP::begin` after the internals of ICP have been readied.
+Example:
 
-\section static_init_sec Static Initialization
+```cpp
+class MyICP final : public icp::ICP {
+public:
+    MyICP();
+    MyICP(const Config& config);
+    ~MyICP();
 
-TODO: update -- we do this in icp.cpp now and the documentation builder will
-automatically search there.
+    void setup() override;    // Optional
+    void iterate() override;
+};
+````
 
-The static initialization is required so that users can instantiate your ICP instance.
-Define
+---
 
-```cpp
-static bool static_initialization = []() {
-    assert(ICP::register_method("name_of_instance", []() -> std::unique_ptr<ICP> {
-        return std::make_unique<NameOfClass>();
-    }));
-    return true;
-}();
-```
+## 3. ICP Lifecycle
 
-where `"name_of_instance"` is the name of your ICP implementation and `NameOfClass` is the name of the class.
+### Provided Instance Variables
 
-\section icp_dpc_sec Documentation
+| Variable    | Type               | Description                        |
+| ----------- | ------------------ | ---------------------------------- |
+| `a`         | `PointCloud`       | Source point cloud                 |
+| `b`         | `PointCloud`       | Target point cloud                 |
+| `match`     | `std::vector<int>` | Matches from `a` to `b`            |
+| `transform` | `RBTransform`      | Current accumulated transformation |
 
-The script icp_doc_builder.py will automatically generate documentation for your ICP instances as markdown files and place them in a desired directory. The invocation format is:
+### Method Responsibilities
 
-```shell
-python3 icp_doc_builder.py dir/where/your/icps/are/ dir/where/markdown/should/go/ where/main/file/is.md
-```
+| Method      | When it is Called            | What to Do                                     |
+| ----------- | --------------------------- | ---------------------------------------------- |
+| `setup()`   | Once before first iteration | Initialize resources (e.g. KD-tree)            |
+| `iterate()` | Once per iteration          | Compute correspondences and update `transform` |
 
-Notably, `where/main/file/is.md` should contain two lines of the form
+---
 
-```md
-<!-- ICP_DOCS_BUILDER EDIT MARKER START -->
-...
-<!-- ICP_DOCS_BUILDER EDIT MARKER END -->
-```
+## 4. Registration
 
-the contents between which markers will be automatically updated by the
-documentation builder.
-
-If the file name is `foo_bar.cpp`, then the Doxygen page reference (from which you can refer to from other pages) will be be `foo_bar_icp Foo_bar`. Information about the file should be encoded in special block comments of the following format.
+To allow users to select your ICP instance by name, add it to the static methods map in `ICP2` or `ICP3` in `icp.cpp`
 
 ```cpp
-/*
-#command values...
-*/
+#include "icp/impl/my_icp.h"
+
+namespace icp {
+    template<>
+    std::unordered_map<std::string, ICPX::MethodConstructor> ICPX::methods{
+        {"my_icp", CONSTRUCT_CONFIG(MyICP)},  // ← Register your method here
+    };
+}
 ```
 
-Supported commands are described below.
-
-- The name of the instance should be given by `/* #name Name of Instance */`.
-- An overview of the algorithm should be provided by `/* #desc Overview of algorithm. */`. This description will be rendered as Doxygen source, so you can use markdown and Doxygen commands such as `\ref`.
-- If your instance uses icp::ICP::Config parameters, document them as `/* #conf "name_of_param" This is a sentence. This is another sentence describing the parameter. */`. This description will be rendered as Doxygen source.
-- Every major step your instance takes should be documented by
-
-    ```cpp
-    /*
-        #step My Step: brief description
-
-        Detailed explanation.
+---
 
-        Sources:
-        https://www.example.com
-        https://www.example.com
-        https://www.example.com
-        https://www.example.com
-    */
-    ```
+## 5. Examples
 
-    The `: brief description` section is optional, as are the detailed explanation and sources sections.
-    These descriptions will be rendered as Doxygen source.
+See the following reference implementations:
 
-See the source code of vanilla.cpp or trimmed.cpp as examples.
+| File                                                     | Type     |
+| -------------------------------------------------------- | ---------|
+| [`vanilla.cpp`](/lib/icp/impl/vanilla.cpp)                | 2d       |
+| [`trimmed.cpp`](/lib/icp/impl/trimmed.cpp)                | 2d       |
+| [`feature_aware.cpp`](/lib/icp/impl/feature_aware.cpp)    | 2d       |
+| [`vanilla_3d.cpp`](/lib/icp/impl/vanilla_3d.cpp)          | 3d       |
+| [`trimmed_3d.cpp`](/lib/icp/impl/trimmed_3d.cpp)          | 3d.      |

include/icp/impl/trimmed_3d.h

@@ -12,7 +12,7 @@
 
 namespace icp {
 
-    class Trimmed3d : public ICP3 {
+    class Trimmed3d final : public ICP3 {
     public:
         Trimmed3d(const Config& config);
         Trimmed3d();

include/icp/impl/vanilla_3d.h

@@ -12,7 +12,7 @@
 
 namespace icp {
 
-    class Vanilla3d : public ICP3 {
+    class Vanilla3d final : public ICP3 {
     public:
         Vanilla3d(const Config& config);
         Vanilla3d();