73cb019a4a
=== 核心变更 ===
- _step_path_resolver 新增 scan_work_dir_for_input() 统一文件扫描工具
- 基于 _SCAN_TABLE 映射表，按 output_type 自动扫描 work_dir 子目录
- 支持扩展名匹配（.dat/.tif/.bsq/.csv 等）和文件名关键词匹配
- 按 mtime 排序，返回最新匹配文件
=== 各面板重构 ===
- step2/3/4/6: 废弃 main_window.stepX_panel 跨面板读取
统一改为 scan_work_dir_for_input(work_dir, 'water_mask/glint_mask/deglint_image')
- step6: 移除对 step1/2/3/5 panel 的4处跨面板依赖
- step8: 替换 _resolve_training_csv_from_workdir 为 scan_work_dir_for_input
优先级：training_spectra_indices -> training_spectra
- step9: 替换 _resolve_latest_wqi_test_csv 为 scan_work_dir_for_input
废弃 factory.get_panel('step4_sampling') 和 get_panel('step8_ml_train')
- step10: 替换3层回退链为 scan_work_dir_for_input('sampling_points')
- step11: 替换 factory.get_panel('step9')/('step1') 为文件扫描
=== 删除的冗余方法 ===
- step8._resolve_training_csv_from_workdir (~55行)
- step9._resolve_latest_wqi_test_csv (~56行)
=== 数据流原则 ===
1. pipeline context 为第一顺位（执行时内存状态）
2. 文件系统扫描为回退（仅信任硬盘上真实存在的文件）
3. 彻底禁止面板间 UI 控件互相读取
274 lines
11 KiB
Python
274 lines
11 KiB
Python
# -*- coding: utf-8 -*-
|
||
"""
|
||
Step 路径解析器——统一消灭 panel 端的硬编码路径与"张冠李戴"跨面板引用。
|
||
|
||
提供三个公共 API：
|
||
- resolve_step_widget(main_window, step_key)
|
||
- get_step_output_path(main_window, step_key, work_dir=None)
|
||
- STEP_DATA_SOURCE 映射表
|
||
|
||
典型使用：
|
||
from src.gui.panels._step_path_resolver import (
|
||
resolve_step_widget, get_step_output_path
|
||
)
|
||
|
||
# 替换前：getattr(main_window.step11_panel, 'output_file', None) # 死代码
|
||
# 替换后：
|
||
widget = resolve_step_widget(main_window, 'step11_predictions') # 找到正确 widget
|
||
if widget: ...
|
||
"""
|
||
|
||
from pathlib import Path
|
||
from typing import Optional, Union
|
||
|
||
|
||
# 用户口语编号 / 业务别名 → PANEL_REGISTRY 中真实 step_id 的映射
|
||
# 这是"张冠李戴"修复的核心——main_window 上已不再直接挂载 panel 属性，
|
||
# 所有面板都通过 _panel_factory.get_panel(step_id) 懒加载访问。
|
||
STEP_DATA_SOURCE = {
|
||
# 数据流 step 编号（用户口语） → PANEL_REGISTRY 中的 step_id
|
||
'step5_clean_output': 'step5_clean',
|
||
'step7_index_output': 'step7_index',
|
||
'step8_ml_train_output': 'step8_ml_train',
|
||
'step8_5_non_empirical': 'step8_ml_train',
|
||
'step9_ml_predict_output': 'step9_ml_predict',
|
||
'step10_watercolor_output': 'step10_watercolor',
|
||
'step11_ml_prediction': 'step9_ml_predict', # 主流程 step11 = ML 预测
|
||
'step12_regression_prediction': 'step8_ml_train', # 主流程 step12 = 非经验预测
|
||
'step13_custom_regression': 'step13_report', # 自定义回归借用 step13 报告面板
|
||
'sampling_csv': 'step4_sampling',
|
||
'training_spectra_csv': 'step5_clean',
|
||
'indices_csv': 'step7_index',
|
||
'models_dir': 'step8_ml_train',
|
||
'watercolor_dir': 'step10_watercolor',
|
||
'prediction_csv_dir': 'step9_ml_predict', # 默认从 ML 预测读
|
||
}
|
||
|
||
|
||
def _read_widget_path(widget) -> str:
|
||
"""统一从 widget 读 path（兼容 FileSelectWidget / QLineEdit / 字符串）。"""
|
||
if widget is None:
|
||
return ""
|
||
if hasattr(widget, 'get_path'):
|
||
try:
|
||
return str(widget.get_path() or "").strip()
|
||
except Exception:
|
||
return ""
|
||
if hasattr(widget, 'text'):
|
||
try:
|
||
return str(widget.text() or "").strip()
|
||
except Exception:
|
||
return ""
|
||
if isinstance(widget, str):
|
||
return widget.strip()
|
||
return ""
|
||
|
||
|
||
def resolve_step_widget(main_window, step_key: str, widget_attr: str = 'output_file'):
|
||
"""通过 panel_factory 访问对应面板，并获取其实际的输入/输出控件。
|
||
|
||
解析顺序：
|
||
1. STEP_DATA_SOURCE[step_key] 找到真实 step_id
|
||
2. main_window._panel_factory.get_panel(step_id) 懒加载拿到面板
|
||
3. getattr(panel, widget_attr) 取出真实控件
|
||
|
||
Returns:
|
||
widget 对象 or None（找不到时返回 None，调用方需自行兜底）
|
||
"""
|
||
step_id = STEP_DATA_SOURCE.get(step_key)
|
||
if not step_id:
|
||
return None
|
||
|
||
# 从主窗口获取工厂，再由工厂通过 step_id 拿出真实面板
|
||
factory = getattr(main_window, '_panel_factory', None)
|
||
if not factory:
|
||
return None
|
||
|
||
panel = factory.get_panel(step_id)
|
||
if not panel:
|
||
return None
|
||
|
||
return getattr(panel, widget_attr, None)
|
||
|
||
|
||
_FALLBACK_DIR_TABLE = {
|
||
# pipeline key（与 _ensure_step_dir_map 对齐）→ 子目录名
|
||
'step1': '1_water_mask',
|
||
'step2': '2_Glint_Detection',
|
||
'step3': '3_deglint',
|
||
'step4_sampling': '4_sampling',
|
||
'step5_clean': '5_Data_Cleaning',
|
||
'step6_feature': '6_Spectral_Feature_Extraction',
|
||
'step7_index': '7_Water_Quality_Indices',
|
||
'step8_ml_train': '8_Supervised_Model_Training',
|
||
'step8': '8_Supervised_Model_Training',
|
||
'step9_ml_predict': '8_Non_Empirical_Regression',
|
||
'step9': '8_Non_Empirical_Regression',
|
||
'step10_watercolor': '10_WaterIndex_Images',
|
||
'step10': '10_WaterIndex_Images',
|
||
'step11_map': '14_visualization',
|
||
'step11': '11_12_13_predictions',
|
||
'step11_predictions': '11_12_13_predictions',
|
||
'step12': '13_Custom_Regression',
|
||
'step12_predictions': '11_12_13_predictions',
|
||
'step13': 'reports',
|
||
'step13_predictions': '11_12_13_predictions',
|
||
'step14': '14_visualization',
|
||
'prediction_dir': '11_12_13_predictions',
|
||
'visualization': '14_visualization',
|
||
'reports': 'reports',
|
||
'custom_regression': '13_Custom_Regression',
|
||
# 扩展：覆盖 panel 内部使用的子目录别名
|
||
'water_mask': '1_water_mask',
|
||
'glint_detection': '2_Glint_Detection',
|
||
'deglint': '3_deglint',
|
||
'sampling': '4_sampling',
|
||
'data_cleaning': '5_Data_Cleaning',
|
||
'spectral_feature': '6_Spectral_Feature_Extraction',
|
||
'indices': '7_Water_Quality_Indices',
|
||
'supervised_models': '8_Supervised_Model_Training',
|
||
'non_empirical': '8_Non_Empirical_Regression',
|
||
'qaa_inversion': '8_QAA_Inversion',
|
||
'regression_modeling': '8_Regression_Modeling',
|
||
'watercolor': '10_WaterIndex_Images',
|
||
'ml_prediction': '9_ML_Prediction',
|
||
'sampling_csv_path': '4_sampling/sampling_spectra.csv',
|
||
}
|
||
|
||
|
||
def get_step_output_path(
|
||
main_window,
|
||
step_key: str,
|
||
work_dir: Optional[Union[str, Path]] = None,
|
||
widget_attr: str = 'output_file',
|
||
fallback_key: Optional[str] = None,
|
||
) -> str:
|
||
"""获取 step_key 指向的输出路径（带 main_window 解析 + 兜底路径）。
|
||
|
||
解析顺序：
|
||
1. STEP_DATA_SOURCE[step_key] 找到对应 panel，从 widget 读用户填的 path
|
||
2. 若为空字符串，用 _FALLBACK_DIR_TABLE[fallback_key or step_key] + work_dir 拼兜底
|
||
3. 全失败返回 str(work_dir)
|
||
|
||
注意：不创建 pipeline 实例（避免触发 osgeo 导入），用本地子目录字典兜底。
|
||
"""
|
||
wd = str(work_dir) if work_dir else ""
|
||
widget = resolve_step_widget(main_window, step_key, widget_attr)
|
||
p = _read_widget_path(widget)
|
||
if p:
|
||
if not Path(p).is_absolute() and wd:
|
||
p = str(Path(wd) / p).replace('\\', '/')
|
||
return p
|
||
|
||
# 兜底：本地子目录字典（与 pipeline._ensure_step_dir_map 一致）
|
||
key = fallback_key or step_key
|
||
sub = _FALLBACK_DIR_TABLE.get(key)
|
||
if sub and wd:
|
||
return str(Path(wd) / sub).replace('\\', '/')
|
||
return wd
|
||
|
||
|
||
def resolve_subdir(work_dir, subdir_key: str) -> str:
|
||
"""纯子目录拼装：把 pipeline key 解析为 work_dir 下的子目录路径。
|
||
|
||
用法：resolve_subdir(self.work_dir, 'visualization')
|
||
→ '<work_dir>/14_visualization'
|
||
|
||
与 pipeline.get_step_output_dir 同源（都查同一份 _FALLBACK_DIR_TABLE 子集）。
|
||
"""
|
||
wd = str(work_dir) if work_dir else ""
|
||
sub = _FALLBACK_DIR_TABLE.get(subdir_key)
|
||
if sub and wd:
|
||
return str(Path(wd) / sub).replace('\\', '/')
|
||
return wd
|
||
|
||
|
||
# ═══════════════════════════════════════════════════════════════
|
||
# 文件系统扫描表 —— 定义每个上游产出类型对应的搜索策略
|
||
# ═══════════════════════════════════════════════════════════════
|
||
_SCAN_TABLE = {
|
||
# output_type → (subdir_key, extensions, file_name_matcher)
|
||
'water_mask': ('water_mask', ['.dat', '.tif', '.tiff'], None),
|
||
'glint_mask': ('glint_detection', ['.dat'], 'severe_glint'),
|
||
'deglint_image': ('deglint', ['.bsq', '.dat', '.tif', '.tiff'], None),
|
||
'sampling_points': ('sampling', ['.csv'], 'sampling_spectra'),
|
||
'processed_data': ('data_cleaning', ['.csv'], 'processed_data'),
|
||
'training_spectra': ('spectral_feature',['.csv'], 'training_spectra'),
|
||
'training_spectra_indices': ('indices', ['.csv'], 'training_spectra_indices'),
|
||
'ml_models_dir': ('supervised_models', None, None), # 目录
|
||
'ml_predictions_dir': ('ml_prediction', None, None), # 目录
|
||
'water_index_csv_dir': ('watercolor', None, None), # 目录
|
||
'visualization_dir': ('visualization', None, None), # 目录
|
||
'reference_img': ('water_mask', ['.bsq', '.dat', '.tif', '.tiff'], None),
|
||
}
|
||
|
||
|
||
def scan_work_dir_for_input(work_dir: str, output_type: str):
|
||
"""基于文件系统扫描查找上游步骤的产出文件。
|
||
|
||
这是 update_from_config 重构的核心工具函数。
|
||
仅信任硬盘上真实存在的文件，彻底消除面板间 UI 控件互相读取。
|
||
|
||
Args:
|
||
work_dir: 工作目录路径
|
||
output_type: 产出类型，如 'water_mask', 'deglint_image', 'sampling_points'
|
||
|
||
Returns:
|
||
找到的文件/目录绝对路径字符串，未找到返回 None
|
||
"""
|
||
if not work_dir:
|
||
return None
|
||
|
||
wd = Path(work_dir)
|
||
entry = _SCAN_TABLE.get(output_type)
|
||
if entry is None:
|
||
return None
|
||
|
||
subdir_key, extensions, matcher = entry
|
||
target_dir = wd / _FALLBACK_DIR_TABLE.get(subdir_key, subdir_key)
|
||
if not target_dir.is_dir():
|
||
return None
|
||
|
||
# 目录类型（extensions 为 None）→ 只要目录存在就返回
|
||
if extensions is None:
|
||
return str(target_dir).replace('\\', '/')
|
||
|
||
# 扫描目录中的文件
|
||
candidates = []
|
||
for ext in extensions:
|
||
for f in target_dir.glob(f'*{ext}'):
|
||
if not f.is_file():
|
||
continue
|
||
candidates.append(f)
|
||
# 也搜 ext 的大写变体
|
||
for f in target_dir.glob(f'*{ext.upper()}'):
|
||
if not f.is_file():
|
||
continue
|
||
candidates.append(f)
|
||
|
||
if not candidates:
|
||
return None
|
||
|
||
# 按 matcher 优先级 + mtime 排序
|
||
if matcher:
|
||
matched = [f for f in candidates if matcher.lower() in f.stem.lower()]
|
||
if matched:
|
||
matched.sort(key=lambda p: p.stat().st_mtime, reverse=True)
|
||
return str(matched[0]).replace('\\', '/')
|
||
# matcher 未命中时，仍返回最新文件（宽松匹配）
|
||
candidates.sort(key=lambda p: p.stat().st_mtime, reverse=True)
|
||
return str(candidates[0]).replace('\\', '/')
|
||
|
||
# 无 matcher → 返回最新的
|
||
candidates.sort(key=lambda p: p.stat().st_mtime, reverse=True)
|
||
return str(candidates[0]).replace('\\', '/')
|
||
|
||
|
||
__all__ = [
|
||
'STEP_DATA_SOURCE',
|
||
'resolve_step_widget',
|
||
'get_step_output_path',
|
||
'resolve_subdir',
|
||
'scan_work_dir_for_input',
|
||
]
|