This skill follows 3coding-visual. It is only responsible for thenon-data diagramsin the paper, such as the technical roadmap, solving flowchart, model structure diagram, data processing flowchart, variable relationship diagram, indicator system diagram, etc.
Mathematical modeling standards reference
If domain judgment is needed, read the "Charts and Visualization" and "Non-data diagram tool selection" subsections in ../_references/math_modeling_norms.md . This file serves only as a standards knowledge base; there is no requirement to generate extra diagrams just to make up the count.
Stage boundaries
- This stage is responsible for: DrawIO source files, non-data diagram PDFs, and diagram generation records.
- This stage is not responsible for: data diagrams such as line charts, bar charts, scatter plots, heatmaps, box plots, and radar charts. These are generated by
3coding-visual. - This stage does not rerun the model, does not modify
code/, and does not rewrite the numerical conclusions ofreports/RESULTS_REPORT.md.
Required outputs
Create or update in the current working directory:
figures/
fig_roadmap.drawio
fig_roadmap.pdf
fig_flow_q1.drawio
fig_flow_q1.pdf
...
reports/DRAWIO_REPORT.md
If a certain type of diagram does not need to be generated, the reason must be explained in reports/DRAWIO_REPORT.md 中说明原因. Competition papers usually need at least one fig_roadmap technical roadmap.
The purpose of reading these files is not to extract data for plotting, but to understand the paper's methods, chapter structure, sub-problem relationships, and existing charts, so as to avoid duplication.
Workflow
Step 1: Take stock of existing charts and requirements
First read the following files (read them if they exist):reports/ANALYSIS_MODELING_REPORT.md、reports/RESULTS_REPORT.md、figures/ directory listing.
Then extract non-data diagram requirements from the preceding documents and output a checklist:
DRAWIO PLAN CHECKLIST:
[ ] fig_roadmap 技术路线图,放在问题重述/绪论
[ ] fig_flow_q1 问题一求解流程图
[ ] fig_flow_q2 问题二求解流程图
[ ] fig_flow_q3 问题三求解流程图
[ ] fig_pipeline 数据处理流程图
[ ] fig_model 模型结构/变量关系图
The checklist is not a fixed template; it should be trimmed or supplemented according to the actual problem. Do not generate meaningless diagrams just to fill in the set.
Step 2: Determine the diagram type
Common diagram choices:
| Diagram type | Suggested file name | Applicable scenario |
|---|---|---|
| Technical roadmap | fig_roadmap |
Show the overall solution route, chapter logic, and method connections |
| Sub-problem solving flowchart | fig_flow_q1, fig_flow_q2 |
Show the input, judgment, algorithm, and output of a single sub-problem |
| Data processing flowchart | fig_pipeline |
Show data cleaning, feature construction, and modeling input |
| Model structure diagram | fig_model |
Show module relationships, variable relationships, and model hierarchy |
| Indicator system diagram | fig_index_system |
Show the goal layer, criterion layer, and indicator layer |
| Decision tree/rule diagram | fig_decision_tree |
Show classification rules, equipment selection, and strategy branches |
Do not use DrawIO to draw these diagrams:
- Result comparison bar chart
- Prediction error curve
- Sensitivity curve
- Correlation heatmap
- Distribution plots and box plots
Step 3: Generate DrawIO source files
One .drawio file per diagram, placed in figures/。
DrawIO content requirements:
- The text language is consistent with the paper language.
- Node text is short, on two lines if necessary, without piling up long sentences.
- Nodes of the same type have consistent styles.
- Arrow directions are clear, avoiding crossings.
- Do not write long explanations in the figure; leave explanations to the main text of the paper.
- Do not use decorative shadows or excessive gradients.
When generating large XML, write it in segments to avoid truncation. Example:
mkdir -p figures
cat << 'XMLEOF' > figures/fig_roadmap.drawio
<mxfile>
<diagram name="Page-1">
<mxGraphModel>
<root>
<mxCell id="0"/>
<mxCell id="1" parent="0"/>
<!-- nodes and edges -->
</root>
</mxGraphModel>
</diagram>
</mxfile>
XMLEOF
Step 4: Export PDF
Prefer using available DrawIO commands to export PDF:
DRAWIO_BIN="$(command -v drawio 2>/dev/null || command -v draw.io 2>/dev/null || command -v draw.io.exe 2>/dev/null || true)"
if [ -n "$DRAWIO_BIN" ]; then
"$DRAWIO_BIN" --export --format pdf --crop --output figures/fig_roadmap.pdf figures/fig_roadmap.drawio
else
echo "DrawIO command not found; keep .drawio source and record export failure."
fi
If PDF cannot be exported, keep .drawio, and in reports/DRAWIO_REPORT.md record the failure reason and suggested export command.
Step 5: Self-check and repair
Each figure must be checked:
.drawiofile is non-empty.- If export succeeds,
.pdffile is non-empty. - Nodes have no obvious overlap.
- Arrows do not pass through core nodes.
- Font size, color, and border style are consistent.
- File names match the meaning of the figures.
- There is no duplication with the data figures in
3coding-visual.
If problems are found, fix .drawio and re-export; do not merely explain in the report.
Step 6: Write the generation record
Create reports/DRAWIO_REPORT.md, including at least:
# DrawIO 图示生成报告
## 图示清单
| 文件 | 类型 | 来源依据 | 用途 | 状态 |
| --- | --- | --- | --- | --- |
## 未生成图示及原因
## 导出与自检记录
## 给论文阶段的嵌入建议
Embedding suggestions only state which chapter each figure is suitable for and the suggested caption; do not generate *_typst_includes.typ. The final figure insertion code (Typst's #figure(image(...), caption: [...]) or LaTeX's \begin{figure}...\end{figure}) is determined by 5writing based on the paper structure and the chosen engine.
Quality requirements
- Figures serve the paper's argument, not decoration.
- Each figure must correspond to a real method in
reports/ANALYSIS_MODELING_REPORT.md. - Data-type charts must not be generated repeatedly at this stage.
- All non-data figures cited in the paper stage should have
.drawiosource files and PDFs, or explain the export failure inreports/DRAWIO_REPORT.md.