Files
playwrite/CLAUDE.md

3.7 KiB

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

This is a Python automation project using Playwright to interact with a Chinese ERP system (YonBIP/用友). The scripts automate login, navigation, form filling, and data submission workflows in a nested iframe environment.

Environment Setup

# Activate the virtual environment (Windows)
.venv\Scripts\activate

# Install dependencies
pip install playwright

# Install Playwright browsers
playwright install chromium

Running Scripts

# Run the main test script
python test_playwright.py

# Run the record script
python record.py

Key Dependencies

  • playwright==1.57.0: Browser automation framework
  • Uses synchronous API (playwright.sync_api)

Architecture

Nested Iframe Structure

The target application uses deeply nested iframes that require careful handling:

page (browser context)
  └── #forwardFrame (main iframe)
      └── #mainiframe (inner iframe, contains actual application UI)

Critical Pattern: When accessing nested iframes, always wait for the inner iframe to be visible before extracting its content frame:

outer_frame = page.locator("#forwardFrame").content_frame
inner_frame_locator = outer_frame.locator("#mainiframe")
inner_frame_locator.wait_for(state="visible", timeout=15000)
inner_frame = inner_frame_locator.content_frame

Loading State Management

The application uses loading overlays ("加载中") that must be detected and waited for:

# Detect and wait for loading to complete
loading_locator = frame.locator("div").filter(has_text="加载中").nth(1)
try:
    loading_locator.wait_for(state="visible", timeout=3000)
    loading_locator.wait_for(state="hidden", timeout=0)  # Infinite wait
except TimeoutError:
    # Loading completed quickly or never appeared
    pass

Page Navigation Pattern

New windows/popups are handled using expect_popup():

with page.expect_popup() as popup_info:
    some_button.click()
new_page = popup_info.value

Common Utilities

get_input_by_label(frame, label_text)

Locates input fields by their associated label text. Searches upward through parent containers to find the input:

input_box = get_input_by_label(frame, "生产部门")
if input_box:
    current_value = input_box.input_value()
    input_box.fill("new value")

click_button_until_disappear(frame, button_name, max_attempts, max_duration)

Continuously clicks a button until it disappears from the DOM. Includes logic to auto-fill missing data before each click.

Development Notes

  • Browser launch: chromium.launch(headless=False) for debugging
  • HTTPS errors: Ignored with ignore_https_errors=True due to self-signed certificates on target system
  • Locators: Uses role-based locators (get_by_role()) where possible, falling back to CSS selectors and text filtering
  • Text matching: Uses regular expressions for matching text with patterns like re.compile(r"^生产部门$")
  • Date format: "YYYY-MM-DD" (e.g., "2025-12-28")
  • Language: Application is in Chinese; comments and print statements use Chinese for clarity
  • Debugging: Uses pdb.set_trace() for interactive debugging sessions

Target Application Details

  • URL: https://68.11.34.30:8082/ (internal network)
  • System: YonBIP (用友) ERP system
  • Login: Requires username and password; may show "force login" confirmation dialog
  • Key workflows:
    • "补货安排" (Replenishment Arrangement)
    • "生产订单" (Production Orders)
    • Batch data submission with "保存提交" (Save & Submit) button