009_collision

G02_collision_box

G01 MCC 截面表加载器回归测试

tests/009_collision/G01_MCC 检查 sub_G01_load_cross_section 的数组边界行为。测试数据是合成的两列数值，不表示真实碰撞截面，也不验证完整 MCC 碰撞物理。

测试内容
输入	验证内容
`cross_section_exact_nmax.dat`	恰好包含 `Nmax` 行；检查数值完整载入且文件结尾探测不越界。
`cross_section_too_many_rows.dat`	包含 `Nmax + 1` 行；检查加载器输出诊断、以非零状态退出，并且 AddressSanitizer 不报告数组越界。

从仓库根目录运行：

bash tests/009_collision/G01_MCC/clean.sh
bash tests/009_collision/G01_MCC/run.sh

成功时会打印两条 PASS 信息。完成后运行：

bash tests/009_collision/G01_MCC/clean.sh

清理命令只删除该测试目录下的 build/。

G02 零维碰撞盒与回归测试

MCC 测试依据与文献对应 C01 概率检查、两体守恒和碰撞盒解析曲线，并区分物理模型验证与软件回归。

库的 C++ 源码和头文件位于 G_Collision/G02_MCC_network/。测试在运行时生成合成 CSV，不携带数据文件或空白表格。用户模型写法见 CSV 数据格式。

初学者先看碰撞盒教程：它观察热粒子与冷气体的碰撞，并解释“零维”、温度与整体运动，以及怎样读结果。下面的回归测试用于检查修改代码后原有功能是否仍正常。

tests/009_collision/G02_MCC_network 独立构建 G02，验证模型包、插值、碰撞算法和 C++ 批量接口；启用 MPI/OpenMP 时还运行并行测试。零维碰撞盒使用合成等质量弹性模型，检查总动量与能量收支、随机碰撞次数、平均速度随时间的变化和末态分布，并包含零密度、零时间步等对照。检查失败时测试返回非零状态。

从仓库根目录运行完整默认测试及手动算例：

bash tests/009_collision/G02_MCC_network/run.sh
bash tests/009_collision/G02_MCC_network/examples/collision_box/run.sh

默认测试需要 C++20 编译器和 CMake 3.20 或以上。 CTest 是自动逐项运行测试的工具；默认脚本还检查内存越界等程序错误。完成时应看到 100% tests passed；手动算例应打印 collision_box: PASS。算例的 checks.csv 中 true 表示该项通过，false 表示失败。

算例将 CSV 和图片写入其已忽略的 results/ 目录。数据仅供软件验证，不得用于科研结论；参数与判据见零维碰撞盒教程。

G01 MCC Cross-Section Loader Regression Test

tests/009_collision/G01_MCC checks the array-bound behavior of sub_G01_load_cross_section. The two-column inputs are synthetic: they are not physical cross sections, and this test does not validate the complete MCC collision model.

Test Cases
Input	What it verifies
`cross_section_exact_nmax.dat`	Contains exactly `Nmax` rows; checks value preservation and a safe end-of-file probe.
`cross_section_too_many_rows.dat`	Contains `Nmax + 1` rows; checks for a diagnostic, a nonzero exit status, and no AddressSanitizer array-bound report.

Run from the repository root:

bash tests/009_collision/G01_MCC/clean.sh
bash tests/009_collision/G01_MCC/run.sh

A successful run prints two PASS messages. Clean the generated executable files with:

bash tests/009_collision/G01_MCC/clean.sh

The clean script removes only this test directory’s build/ directory.

G02 Collision-Box and Regression Tests

MCC verification references map the C01 probability checks, binary conservation and analytic box curves to their bases, distinguishing model verification from software regression.

Library C++ sources/headers are in G_Collision/G02_MCC_network/. Tests generate synthetic CSV at runtime; no data files or blank tables are shipped. See CSV format for user input.

Start with the collision-box tutorial for hot particles in a cold gas, zero-dimensional modeling, temperature versus drift, and reading the output. Regression tests check that existing behavior still works after code changes.

tests/009_collision/G02_MCC_network independently builds G02 and tests model packages, interpolation, collision algorithms and C++ batching and optional MPI/OpenMP paths. The zero-dimensional collision box in G02_collision_box uses a synthetic equal-mass elastic model. It checks conservation, Poisson collision counts, the time evolution of mean velocities, and the final distribution, with zero-density and zero-time-step controls. A failed check returns a nonzero status.

From the repository root, run the default suite and the manual example:

bash tests/009_collision/G02_MCC_network/run.sh
bash tests/009_collision/G02_MCC_network/examples/collision_box/run.sh

The default tests need a C++20 compiler and CMake 3.20 or newer. CTest runs each test automatically; the default script also checks memory errors. Look for 100% tests passed and, for the manual example, collision_box: PASS. In its checks.csv, true means a check passed and false means it failed.

The example writes CSV files and plots to its ignored results/ directory. Its data are for software validation only, not scientific conclusions; see the collision-box tutorial for parameters and pass criteria.