airmosaic-ai

Industrial yearbook tables

English 中文

Use existing Excel tables to prepare province–industry economic indicators or province–product output. National-sector processing requires a separate implementation.

Prepare a task copy

Use R with data.table, readxl, and stringr, plus a spreadsheet viewer for source checks. Record edition years, actual data years, authorized files/sheets, units and industry-code version. Put immutable originals in an external versioned data folder; create the task specification and mapping in AIRMOSAIC_CONFIG_DIR, with working copies and results in a fresh project result folder.

The two R templates are adaptation material: yearbook_regional_sector_template and yearbook_product_output_template cannot execute through run-tool. Inspect their current metadata without starting processing:

python -m runtime.cli describe-tool yearbook_regional_sector_template
python -m runtime.cli describe-tool yearbook_product_output_template

Adapt and run only the reviewed task script

  1. Inspect representative authorized sheets in a spreadsheet viewer. Regional tables are expected beneath 中国工业统计年鉴{edition}（Excel）/3分地区数据/; product tables beneath 工业产量/ or 分地区工业主要产品产量/ in each edition folder.
  2. Copy the required regional or product template into the task workspace and review it before sourcing. The current files contain incomplete/invalid expressions and simplified handling; they are not ready-to-run recipes. Validate syntax and repair the task copy separately before testing on synthetic sheets.
  3. Set explicit input files, edition range, source-sheet selection, mappings and a fresh output folder in the reviewed task copy. config.yaml is a declaration only: templates neither load it nor expand its placeholders. Existing examples are structural illustrations.
  4. Review every write: regional per-year processing writes inside edition folders despite its final output_dir; product processing writes within edition folders and root, with no independent output-directory argument. Use task working copies or adapt all writes before any use; setting one final destination does not protect originals.
  5. Check extracted headers and units, 地区 panel boundaries, continuation sheets, numeric-conversion failures, edition/data-year mapping, excluded tables and industry-code matches. The templates assume data year = edition year − 1; verify that assumption against actual tables.
  6. Export the cleaned table, source/edition inventory, mappings, conversion issues, missing coverage and all fill rules. Independently check several source cells and year/region/industry/product totals. Missing values are not automatically zeros; optional interpolation, extrapolation or supplementation needs a separately justified task implementation.

Detailed layout and historical design: processing techniques, pipeline blueprint, industry matching, encoding repair. Historical references retain their existing language. Scientific choices and interpretation: yearbook methods.

This site is open source. Improve this page.