Skip to content

An open-source, privacy-first, self-hosted knowledge workspace where humans and AI agents work together 开源、隐私优先、自托管的知识工作空间,让人与智能体在此协作

17.2 KB
Raw Blame
// SiYuan - https://github.com/siyuan-note/siyuan
// Copyright (c) 2020-present, b3log.org
// SPDX-License-Identifier: AGPL-3.0-or-later

package api

import (
	"github.com/siyuan-note/siyuan/kernel/apicontract"
	"html"
	"net/http"
	"os"
	"path/filepath"
	"reflect"
	"strings"

	"github.com/88250/gulu"
	"github.com/88250/lute"
	"github.com/88250/lute/ast"
	"github.com/88250/lute/parse"
	"github.com/88250/lute/render"
	"github.com/PuerkitoBio/goquery"
	"github.com/gin-gonic/gin"
	"github.com/siyuan-note/logging"
	"github.com/siyuan-note/siyuan/kernel/model"
	"github.com/siyuan-note/siyuan/kernel/treenode"
	"github.com/siyuan-note/siyuan/kernel/util"
)

// maxSpinBlockDOMBytes 限制 spinBlockDOM 输入 DOM 的最大字节数。
const maxSpinBlockDOMBytes = 1024 * 1024

// maxHTML2BlockDOMRequestBytes 限制 html2BlockDOM 请求体的最大字节数。
const maxHTML2BlockDOMRequestBytes int64 = 128 * 1024 * 1024

var copyStdMarkdown = contractHandler(apicontract.CopyStdMarkdown, func(c *gin.Context, request apicontract.CopyStdMarkdownRequest) apicontract.Response[string] {

	id := request.ID
	assetsDestSpace2Underscore, fillCSSVar, adjustHeadingLevel, imgTag := request.AssetsDestSpace2Underscore, request.FillCSSVar, request.AdjustHeadingLevel, request.ImgTag
	isReadOnlyRole := model.IsReadOnlyRoleContext(c)
	avPublishFilter := model.NewAVExportPublishFilter(c)
	var publishAccess model.PublishAccess
	var accessChecker model.EmbedBlockAccessChecker
	if isReadOnlyRole {
		publishAccess = model.GetPublishAccess()
		accessChecker = func(blockID string) bool {
			return model.CheckBlockIdAccessableByPublishAccess(c, publishAccess, blockID)
		}
	}
	markdownContent := model.ExportStdMarkdown(id, &model.ExportOptions{
		Render: model.ExportRenderOptions{
			AssetsDestSpace2Underscore: assetsDestSpace2Underscore,
			FillCSSVar:                 fillCSSVar,
			AdjustHeadingLevel:         adjustHeadingLevel,
			ImgTag:                     imgTag,
			AVPublishFilter:            avPublishFilter,
			AccessCheckers:             []model.EmbedBlockAccessChecker{accessChecker},
		},
	})
	if isReadOnlyRole {
		bt := treenode.GetBlockTree(id)
		if bt != nil {
			markdownContent = model.FilterContentByPublishAccess(c, publishAccess, bt.BoxID, bt.Path, markdownContent, true)
		}
	}
	return apicontract.Success(markdownContent)
})

var html2BlockDOM = contractHandler(apicontract.HTML2BlockDOM, func(c *gin.Context, request apicontract.HTMLClipboardRequest) apicontract.Response[apicontract.HTMLClipboardData] {
	dom := request.DOM
	// 可选 notebook 参数:指定目标加密笔记本时资源写入 box 内并加密
	boxID := ""
	if notebook := request.Notebook; notebook != "" {
		if model.IsEncryptedBox(notebook) {
			boxID = notebook
		}
	}
	if err := holdEncryptedBoxRequest(c, boxID); err != nil {
		return apicontract.Failure[apicontract.HTMLClipboardData](-1, err.Error())
	}
	text := request.Text
	mathML := request.MathML
	office := request.Office
	officeMathHTML := request.OfficeMathHTML
	wps := request.WPS
	skipLocalAssets := request.SkipLocalAssets
	skipBase64Assets := request.SkipBase64Assets
	skipInlineSVGAssets := request.SkipInlineSVGAssets
	preflight := request.Preflight
	preparedHTML := request.PreparedHTML
	preserveSourceFormat := request.PreserveSourceFormat
	luteEngine := util.NewLute()
	luteEngine.SetHTMLTag2TextMark(true)
	luteEngine.SetHTML2MarkdownAttrs([]string{"alias", "memo", "bookmark", "custom-*"})
	luteEngine.SetLatexMath(model.MathDelimiterCompatibilityEnabled())
	dom, useHTML := prepareHTMLClipboardContent(luteEngine, dom, text, mathML, office, officeMathHTML, wps,
		preserveSourceFormat, preparedHTML, convertClipboardMath, convertOfficeHTMLClipboardMath)
	if !useHTML {
		if preflight {
			return apicontract.Success(apicontract.HTMLClipboardPrepared(apicontract.HTMLClipboardPreflight{Converted: true, DOM: &dom, UseHTML: false}))
		} else {
			return apicontract.Success(apicontract.HTMLClipboardText(dom))
		}
	}
	if preflight {
		skipLocalAssets = true
		skipBase64Assets = true
		skipInlineSVGAssets = true
	}
	normalizedHTML := dom
	tree, _ := model.HTML2TreeWithOptions(dom, luteEngine, boxID, model.HTML2TreeOptions{
		SkipBase64Assets:    skipBase64Assets,
		SkipInlineSVGAssets: skipInlineSVGAssets,
	})
	if nil == tree {
		if preflight {
			failedDOM := "Failed to convert"
			return apicontract.Success(apicontract.HTMLClipboardPrepared(apicontract.HTMLClipboardPreflight{DOM: &failedDOM, NormalizedHTML: &normalizedHTML, UseHTML: true}))
		} else {
			return apicontract.Success(apicontract.HTMLClipboardText("Failed to convert"))
		}
	}

	var unlinks []*ast.Node
	ast.Walk(tree.Root, func(n *ast.Node, entering bool) ast.WalkStatus {
		if !entering {
			return ast.WalkContinue
		}

		if ast.NodeListItem == n.Type && nil == n.FirstChild {
			newNode := treenode.NewParagraph("")
			n.AppendChild(newNode)
			n.SetIALAttr("updated", util.TimeFromID(newNode.ID))
			return ast.WalkSkipChildren
		} else if ast.NodeBlockquote == n.Type && nil == n.FirstChild.Next {
			unlinks = append(unlinks, n)
		}
		return ast.WalkContinue
	})
	for _, n := range unlinks {
		n.Unlink()
	}

	// 表格只包含一个单元格时,将其转换为段落
	// Copy one cell from Excel/HTML table and paste it using the cell's content https://github.com/siyuan-note/siyuan/issues/9614
	unlinks = nil
	if nil != tree.Root.FirstChild && ast.NodeTable == tree.Root.FirstChild.Type && (nil == tree.Root.FirstChild.Next ||
		(ast.NodeKramdownBlockIAL == tree.Root.FirstChild.Next.Type && nil == tree.Root.FirstChild.Next.Next)) {
		if nil != tree.Root.FirstChild.FirstChild && ast.NodeTableHead == tree.Root.FirstChild.FirstChild.Type {
			head := tree.Root.FirstChild.FirstChild
			if nil == head.Next && nil != head.FirstChild && nil == head.FirstChild.Next {
				row := head.FirstChild
				if nil != row.FirstChild && nil == row.FirstChild.Next {
					cell := row.FirstChild
					p := treenode.NewParagraph("")
					var contents []*ast.Node
					for c := cell.FirstChild; nil != c; c = c.Next {
						contents = append(contents, c)
					}
					for _, c := range contents {
						p.AppendChild(c)
					}
					tree.Root.FirstChild.Unlink()
					tree.Root.PrependChild(p)
				}
			}
		}
	}

	if util.ContainerStd == model.Conf.System.Load().Container && !skipLocalAssets {
		// 处理本地资源文件复制
		ast.Walk(tree.Root, func(n *ast.Node, entering bool) ast.WalkStatus {
			if !entering || ast.NodeLinkDest != n.Type {
				return ast.WalkContinue
			}

			if "" == n.TokensStr() {
				return ast.WalkContinue
			}

			localPath := n.TokensStr()
			if strings.HasPrefix(localPath, "http") {
				return ast.WalkContinue
			}
			localPath = util.FileURLToLocalPath(localPath)
			if !filepath.IsAbs(localPath) {
				// Kernel crash when copy-pasting from some browsers https://github.com/siyuan-note/siyuan/issues/9203
				return ast.WalkContinue
			}
			if !gulu.File.IsExist(localPath) {
				return ast.WalkContinue
			}

			if util.IsSensitivePath(localPath) {
				logging.LogWarnf("skip copying asset [%s] due to sensitive path", localPath)
				return ast.WalkContinue
			}
			if encryptedBoxID := model.EncryptedRawPathBoxID(localPath); encryptedBoxID != "" {
				logging.LogWarnf("skip copying asset [%s] from encrypted notebook [%s]", localPath, encryptedBoxID)
				return ast.WalkContinue
			}

			name := filepath.Base(localPath)
			ext := filepath.Ext(name)
			name = name[0 : len(name)-len(ext)]
			name = name + "-" + ast.NewNodeID() + ext

			data, readErr := os.ReadFile(localPath)
			if readErr != nil {
				logging.LogErrorf("read asset [%s] failed: %s", localPath, readErr)
				return ast.WalkStop
			}
			assetsDir := filepath.Join(util.DataDir, "assets")
			if boxID != "" {
				assetsDir = filepath.Join(util.DataDir, boxID, "assets")
			}
			storedName, storeErr := model.StoreAssetForBox(boxID, assetsDir, name, data)
			if storeErr != nil {
				logging.LogErrorf("store asset [%s] failed: %s", localPath, storeErr)
				return ast.WalkStop
			}
			assetURL := "assets/" + storedName
			if boxID != "" {
				assetURL += "?box=" + boxID
			}
			n.Tokens = gulu.Str.ToBytes(assetURL)
			return ast.WalkContinue
		})
	}

	parse.TextMarks2Inlines(tree) // 先将 TextMark 转换为 Inlines https://github.com/siyuan-note/siyuan/issues/13056
	parse.NestedInlines2FlattedSpansHybrid(tree, false)
	// 合并转义节点拆分出的同格式文本,避免重解析时将片段边缘空白重复移到标记外。
	ast.Walk(tree.Root, func(n *ast.Node, entering bool) ast.WalkStatus {
		if entering && n.Type == ast.NodeTextMark && n.Previous != nil && n.Previous.Type == ast.NodeTextMark &&
			n.ContainTextMarkTypes("strong", "em", "s", "mark", "sup", "sub") &&
			n.TextMarkATitle == n.Previous.TextMarkATitle && reflect.DeepEqual(n.KramdownIAL, n.Previous.KramdownIAL) {
			luteEngine.MergeSameTextMark(n)
		}
		return ast.WalkContinue
	})
	removeWhitespaceTextMarkStyles(tree)

	md, err := lute.FormatNodeSync(tree.Root, luteEngine.ParseOptions, luteEngine.RenderOptions)
	if nil != err {
		if preflight {
			failedDOM := "Failed to convert"
			return apicontract.Success(apicontract.HTMLClipboardPrepared(apicontract.HTMLClipboardPreflight{DOM: &failedDOM, NormalizedHTML: &normalizedHTML, UseHTML: true}))
		} else {
			return apicontract.Success(apicontract.HTMLClipboardText("Failed to convert"))
		}
	}

	// 中间格式已经对 HTML 文本编码,重解析时保留实体,避免将正文中的标签当作 DOM 渲染。
	luteEngine.ParseOptions.KeepEscaped = true
	// HTML 入口已完成公式归一化,避免再次解释链接等受保护区域中的替代分隔符。
	luteEngine.SetLatexMath(false)
	tree = parse.Parse("", []byte(md), luteEngine.ParseOptions)
	ast.Walk(tree.Root, func(n *ast.Node, entering bool) ast.WalkStatus {
		if entering && ast.NodeIFrame == n.Type {
			normalizeIFramePosition(n)
		}
		return ast.WalkContinue
	})
	renderer := render.NewProtyleRenderer(tree, luteEngine.RenderOptions, luteEngine.ParseOptions)
	output := renderer.Render()
	if preflight {
		return apicontract.Success(apicontract.HTMLClipboardPrepared(apicontract.HTMLClipboardPreflight{Converted: true, NormalizedHTML: &normalizedHTML, UseHTML: true}))
	} else {
		return apicontract.Success(apicontract.HTMLClipboardText(gulu.Str.FromBytes(output)))
	}
}, func(c *gin.Context) *apicontract.Response[apicontract.HTMLClipboardData] {
	limitHTML2BlockDOMRequestBody(c, maxHTML2BlockDOMRequestBytes)
	return nil
})

func removeWhitespaceTextMarkStyles(tree *parse.Tree) {
	var unlinks []*ast.Node
	ast.Walk(tree.Root, func(n *ast.Node, entering bool) ast.WalkStatus {
		if entering && n.Type == ast.NodeTextMark && strings.TrimSpace(n.TextMarkTextContent) == "" {
			n.RemoveIALAttr("style")
			// 空白文本无法绑定 Kramdown 行内属性,移除样式属性节点以免渲染为正文。
			if n.Next != nil && n.Next.Type == ast.NodeKramdownSpanIAL && parse.IALVal(n.Next, "style") != "" {
				unlinks = append(unlinks, n.Next)
			}
		}
		return ast.WalkContinue
	})
	for _, n := range unlinks {
		n.Unlink()
	}
}

func normalizeIFramePosition(node *ast.Node) {
	// iframe 的绝对或固定定位依赖原网页容器,粘贴后需让块保留在文档流中。
	if isOutOfFlowPosition(node.IALAttr("style")) {
		node.RemoveIALAttr("style")
	}

	doc, err := goquery.NewDocumentFromReader(strings.NewReader(node.TokensStr()))
	if err != nil {
		return
	}
	iframe := doc.Find("iframe").First()
	style, exists := iframe.Attr("style")
	if !exists || !isOutOfFlowPosition(style) {
		return
	}

	iframe.RemoveAttr("style")
	if normalized, renderErr := goquery.OuterHtml(iframe); renderErr == nil {
		node.Tokens = []byte(normalized)
	}
}

func isOutOfFlowPosition(style string) bool {
	position := ""
	for declaration := range strings.SplitSeq(style, ";") {
		property, value, ok := strings.Cut(declaration, ":")
		if !ok || !strings.EqualFold(strings.TrimSpace(property), "position") {
			continue
		}
		value = strings.ToLower(strings.TrimSpace(value))
		value = strings.TrimSpace(strings.TrimSuffix(value, "!important"))
		position = value
	}
	return position == "absolute" || position == "fixed"
}

func limitHTML2BlockDOMRequestBody(c *gin.Context, maxBytes int64) {
	c.Request.Body = http.MaxBytesReader(c.Writer, c.Request.Body, maxBytes)
}

type clipboardMathConverter func(mathML, office, wps string) (markdown string, converted bool)
type officeHTMLClipboardMathConverter func(officeMathHTML string) (markdown string, converted bool)

func prepareHTMLClipboardContent(luteEngine *lute.Lute, dom, text, mathML, office, officeMathHTML, wps string,
	preserveSourceFormat, preparedHTML bool, mathConverter clipboardMathConverter,
	officeHTMLMathConverter officeHTMLClipboardMathConverter) (resolvedDOM string, useHTML bool) {
	if preparedHTML {
		return dom, true
	}
	resolvedDOM, useHTML = resolveHTMLClipboardContent(luteEngine, dom, mathML, office, officeMathHTML, wps,
		mathConverter, officeHTMLMathConverter)
	if !useHTML {
		return resolvedDOM, false
	}
	// 将 Word 和 WPS 批注转换为行级备注 https://github.com/siyuan-note/siyuan/issues/18748
	resolvedDOM = normalizeWPSComments(resolvedDOM, text, wps)
	resolvedDOM = normalizeMSWordComments(resolvedDOM)
	if !preserveSourceFormat {
		resolvedDOM = matchHTMLClipboardElements(resolvedDOM)
	}
	return resolvedDOM, true
}

func resolveHTMLClipboardContent(luteEngine *lute.Lute, dom, mathML, office, officeMathHTML, wps string,
	mathConverter clipboardMathConverter, officeHTMLMathConverter officeHTMLClipboardMathConverter) (resolvedDOM string, useHTML bool) {
	// 将 Word 和 WPS 公式转换为可编辑公式 https://github.com/siyuan-note/siyuan/issues/18747
	if markdown, converted := mathConverter(mathML, office, wps); converted {
		luteEngine.SetInlineMath(true)
		return luteEngine.Md2BlockDOM(markdown, false), false
	}
	if markdown, converted := officeHTMLMathConverter(officeMathHTML); converted {
		luteEngine.SetInlineMath(true)
		return luteEngine.Md2BlockDOM(markdown, false), false
	}
	return dom, true
}

func normalizeMSWordComments(dom string) string {
	if !strings.Contains(dom, "mso-comment") && !strings.Contains(dom, "MsoComment") && !strings.Contains(dom, "msocom") {
		return dom
	}

	doc, err := goquery.NewDocumentFromReader(strings.NewReader(dom))
	if err != nil {
		return dom
	}

	comments := map[string]string{}
	doc.Find(`[id^="_com_"]`).Each(func(_ int, comment *goquery.Selection) {
		id, _ := comment.Attr("id")
		var paragraphs []string
		comment.Find(".MsoCommentText").Each(func(_ int, paragraph *goquery.Selection) {
			paragraph = paragraph.Clone()
			paragraph.Find(".MsoCommentReference, .msocomoff").Remove()
			if text := strings.TrimSpace(paragraph.Text()); text != "" {
				paragraphs = append(paragraphs, text)
			}
		})
		if 0 < len(paragraphs) {
			comments[strings.TrimPrefix(id, "_com_")] = strings.Join(paragraphs, "\n")
		}
	})

	doc.Find("a[style]").Each(func(_ int, anchor *goquery.Selection) {
		style, _ := anchor.Attr("style")
		if !strings.Contains(strings.ToLower(style), "mso-comment-reference:") {
			return
		}
		if href, exists := anchor.Attr("href"); exists && href != "" {
			return
		}

		content, err := anchor.Html()
		if err != nil {
			return
		}
		comment := comments[msWordCommentID(anchor, style)]
		if comment == "" {
			anchor.ReplaceWithHtml(content)
			return
		}
		anchor.ReplaceWithHtml(`<span title="` + html.EscapeString(comment) + `">` + content + `</span>`)
	})

	doc.Find(".MsoCommentReference, .msocomanchor, .msocomoff").Remove()
	doc.Find("[style]").Each(func(_ int, selection *goquery.Selection) {
		style, _ := selection.Attr("style")
		if strings.Contains(strings.ToLower(style), "mso-element:comment-list") {
			selection.Remove()
		}
	})

	ret, err := doc.Find("body").Html()
	if err != nil {
		return dom
	}
	return ret
}

func msWordCommentID(anchor *goquery.Selection, style string) string {
	if href, exists := anchor.Next().Find(`a[href^="#_msocom_"]`).First().Attr("href"); exists {
		return strings.TrimPrefix(href, "#_msocom_")
	}

	style = style[strings.Index(strings.ToLower(style), "mso-comment-reference:")+len("mso-comment-reference:"):]
	if semicolon := strings.IndexByte(style, ';'); 0 <= semicolon {
		style = style[:semicolon]
	}
	if underscore := strings.LastIndexByte(style, '_'); 0 <= underscore {
		return strings.TrimSpace(style[underscore+1:])
	}
	return ""
}

var spinBlockDOM = contractHandler(apicontract.SpinBlockDOM, func(c *gin.Context, request apicontract.DOMTextRequest) apicontract.Response[apicontract.DOMData] {

	dom := request.DOM
	if len(dom) > maxSpinBlockDOMBytes {
		// 限制输入大小,避免解析超大 DOM 导致资源消耗
		return apicontract.Failure[apicontract.DOMData](http.StatusRequestEntityTooLarge, "dom input exceeds the maximum permitted size")
	}
	luteEngine := model.NewLute()

	dom = luteEngine.SpinBlockDOM(dom)
	return apicontract.Success(apicontract.DOMData{DOM: dom})
})

// md2HTML 将 Markdown 转换为 HTML。
var md2HTML = contractHandler(apicontract.Md2HTML, func(c *gin.Context, request apicontract.MarkdownHTMLRequest) apicontract.Response[apicontract.HTMLData] {

	markdown, mode := request.Markdown, request.Mode

	var html string
	switch mode {
	case "protyle-preview":
		html = model.MarkdownToProtylePreviewHTML(markdown)
	case "":
		html = model.MarkdownToMarkdownStrHTML(markdown)
	default:
		return apicontract.Failure[apicontract.HTMLData](-1, "unknown [mode]")
	}

	return apicontract.Success(apicontract.HTMLData{HTML: html})
})