schedule_converter/parser/cet.go

package parser

import (
	"fmt"
	"regexp"
	"strings"

	"github.com/PuerkitoBio/goquery"
)

// CETScore 四六级成绩条目。
type CETScore struct {
	Level     string // 等级：英语四级/英语六级
	ExamTime  string // 考试时间：如 2025年06月
	TicketNum string // 准考证号
	Score     string // 成绩
}

// CETStudentInfo 四六级查询中的学生信息。
type CETStudentInfo struct {
	StudentID string // 学号
	Name      string // 姓名
	Gender    string // 性别
	IDCard    string // 身份证号
	Duration  string // 学制
	Grade     string // 年级
	College   string // 学院
	Major     string // 专业
}

// CETResult 四六级查询结果。
type CETResult struct {
	StudentInfo CETStudentInfo
	Scores      []CETScore
}

// cetScoreRegex 匹配四六级成绩行。
// 示例：英语六级  时间 2025年06月 准考证号  372220251213105 成绩  445
var cetScoreRegex = regexp.MustCompile(`(英语[四六]级)\s+时间\s*([0-9]{4}年[0-9]{1,2}月)\s+准考证号\s*([0-9A-Za-z]+)\s+成绩\s*([0-9]+|缺考|合格|不合格|优秀|良好|通过|未通过)`)

// FetchCETHTML 解析四六级成绩 HTML 页面。
func ParseCETHTML(html string) (*CETResult, error) {
	doc, err := goquery.NewDocumentFromReader(strings.NewReader(html))
	if err != nil {
		return nil, err
	}

	result := &CETResult{}
	table := findCETTable(doc)
	if table.Length() == 0 {
		return nil, fmt.Errorf("未找到四六级成绩表格")
	}

	parseCETStudentInfo(table, &result.StudentInfo)
	result.Scores = parseCETScores(table)
	if len(result.Scores) == 0 {
		result.Scores = parseCETScoresFromText(doc.Find(".jdcx_area").Text())
	}
	if len(result.Scores) == 0 {
		result.Scores = parseCETScoresFromText(doc.Find(".Contentbox").Text())
	}
	if len(result.Scores) == 0 {
		result.Scores = parseCETScoresFromText(doc.Text())
	}

	return result, nil
}

func findCETTable(doc *goquery.Document) *goquery.Selection {
	selectors := []string{
		"table.addlist",
		".jdcx_area table",
		"form#ff table",
		"table:contains('已有成绩')",
	}
	for _, selector := range selectors {
		table := doc.Find(selector).First()
		if table.Length() > 0 {
			return table
		}
	}
	return doc.Find("__not_found__")
}

func parseCETStudentInfo(table *goquery.Selection, info *CETStudentInfo) {
	table.Find("tr").Each(func(_ int, row *goquery.Selection) {
		var label string
		row.Find("th, td").Each(func(_ int, cell *goquery.Selection) {
			text := normalizeCETText(cell.Text())
			if text == "" {
				return
			}

			if cell.Is("th") {
				label = normalizeCETLabel(text)
				return
			}

			assignCETStudentField(info, label, text)
			label = ""
		})
	})
}

func assignCETStudentField(info *CETStudentInfo, label, value string) {
	switch label {
	case "学号":
		info.StudentID = value
	case "姓名":
		info.Name = value
	case "性别":
		info.Gender = value
	case "身份证号":
		info.IDCard = value
	case "学制":
		info.Duration = value
	case "年级":
		info.Grade = value
	case "学院":
		info.College = value
	case "专业":
		info.Major = value
	}
}

func parseCETScores(table *goquery.Selection) []CETScore {
	var scores []CETScore
	table.Find("tr").Each(func(_ int, row *goquery.Selection) {
		rowText := normalizeCETText(row.Text())
		if !strings.Contains(rowText, "英语四级") && !strings.Contains(rowText, "英语六级") {
			return
		}
		scores = append(scores, parseCETScoresFromText(rowText)...)
	})
	return deduplicateCETScores(scores)
}

func parseCETScoresFromText(text string) []CETScore {
	text = normalizeCETText(text)
	matches := cetScoreRegex.FindAllStringSubmatch(text, -1)
	scores := make([]CETScore, 0, len(matches))
	for _, match := range matches {
		if len(match) < 5 {
			continue
		}
		scores = append(scores, CETScore{
			Level:     strings.TrimSpace(match[1]),
			ExamTime:  strings.TrimSpace(match[2]),
			TicketNum: strings.TrimSpace(match[3]),
			Score:     strings.TrimSpace(match[4]),
		})
	}
	return deduplicateCETScores(scores)
}

func deduplicateCETScores(scores []CETScore) []CETScore {
	if len(scores) <= 1 {
		return scores
	}
	seen := make(map[string]struct{}, len(scores))
	result := make([]CETScore, 0, len(scores))
	for _, score := range scores {
		key := strings.Join([]string{score.Level, score.ExamTime, score.TicketNum, score.Score}, "|")
		if _, ok := seen[key]; ok {
			continue
		}
		seen[key] = struct{}{}
		result = append(result, score)
	}
	return result
}

func normalizeCETLabel(label string) string {
	label = normalizeCETText(label)
	label = strings.ReplaceAll(label, " ", "")
	label = strings.ReplaceAll(label, "：", "")
	label = strings.ReplaceAll(label, ":", "")
	return label
}

func normalizeCETText(text string) string {
	text = strings.ReplaceAll(text, "\u00a0", " ")
	return strings.Join(strings.Fields(strings.TrimSpace(text)), " ")
}