blog/parse.go

// parse.go
package main

import (
	"bytes"
	"fmt"
	"os"
	"path/filepath"
	"regexp"
	"strings"
	"time"

	"git.else-if.org/jess/blog/templates"
	"github.com/BurntSushi/toml"
	mathjax "github.com/litao91/goldmark-mathjax"
	"github.com/yuin/goldmark"
	"github.com/yuin/goldmark/extension"
	"github.com/yuin/goldmark/parser"
	"github.com/yuin/goldmark/renderer/html"
	"go.abhg.dev/goldmark/frontmatter"
	"gopkg.in/yaml.v3"
)

// GetContentPath ensures the content is cached and returns the path to the cached file.
func GetContentPath(file ContentFile) (string, error) {
	cachePath := GetCacheFilename(file.OriginalPath)



	// Read Raw File
	raw, err := ReadRaw(file.OriginalPath)
	if err != nil {
		return "", fmt.Errorf("failed to read file: %w", err)
	}

	var dataToWrite []byte

	if !file.IsMarkdown {
		dataToWrite = raw
	} else {
		// Configure Goldmark with the custom tag extension
		md := goldmark.New(
			goldmark.WithExtensions(
				extension.GFM,
				&frontmatter.Extender{},
				mathjax.MathJax,
				templates.SidebarTag, // ||| -> <sidebar>
				templates.TopBanner,
			),
			goldmark.WithRendererOptions(
				html.WithUnsafe(),
			),
		)

		var buf bytes.Buffer
		ctx := parser.NewContext()
		if err := md.Convert(raw, &buf, parser.WithContext(ctx)); err != nil {
			return "", fmt.Errorf("failed to parse markdown: %w", err)
		}

		// Extract Frontmatter
		// 1. Initialize with External Metadata (if any)
		var meta templates.PageMetadata
		if err := loadExternalMetadata(file.OriginalPath, &meta); err != nil {
			fmt.Printf("Warning: failed to load external metadata for %s: %v\n", file.OriginalPath, err)
		}

		// 2. Parse Internal Frontmatter (overrides external)
		d := frontmatter.Get(ctx)
		if d != nil {
			// Decode into a temporary map to capture all fields
			var rawMap map[string]interface{}
			if err := d.Decode(&rawMap); err != nil {
				fmt.Printf("Warning: failed to decode frontmatter for %s: %v\n", file.OriginalPath, err)
			} else {
				// Merge into meta
				if meta.Raw == nil {
					meta.Raw = make(map[string]interface{})
				}
				for k, v := range rawMap {
					meta.Raw[k] = v
					// Also update struct fields manually if needed, or use layout decoding
					switch strings.ToLower(k) {
					case "title":
						if s, ok := v.(string); ok {
							meta.Title = s
						}
					case "stylesheet":
						if s, ok := v.(string); ok {
							meta.Stylesheet = s
						}
					case "style":
						if s, ok := v.(string); ok {
							meta.Style = s
						}
					}
				}
			}
		}

		// 3. Resolve Stylesheet Path
		if meta.Stylesheet != "" {
			// If not absolute path (starting with /), resolve relative to content file
			if !strings.HasPrefix(meta.Stylesheet, "/") {
				// Get valid web path for the directory containing the file
				// file.RoutePath is like "/entries/entry1" or "/index"
				// We want the directory part of the RoutePath
				
				dirRoute := filepath.Dir(file.RoutePath)
				if dirRoute == "." {
					dirRoute = ""
				}
				
				// Construct new path
				// e.g. /entries + / + style.css -> /entries/style.css
				// Clean handles double slashes
				meta.Stylesheet = filepath.Join(dirRoute, meta.Stylesheet)
				
				// Ensure it starts with / for web usage
				if !strings.HasPrefix(meta.Stylesheet, "/") {
					meta.Stylesheet = "/" + meta.Stylesheet
				}
			}
		}

		if meta.Title == "" {
			base := filepath.Base(file.OriginalPath)
			meta.Title = strings.TrimSuffix(base, filepath.Ext(base))
		}

		htmlContent := buf.String()

		// 3. Post-Process for Index (Dynamic Content Injection)
		// If Home Page, replace content with Latest Post Preview
		if file.RoutePath == "/" {
			// Find latest post
			var latestPost *ContentFile
			for _, f := range AllContent {
				if f.IsMarkdown && f.RoutePath != "/" {
					// Found one
					latestPost = &f
					break // AllContent is sorted by time descending
				}
			}
			
			if latestPost != nil {
				// Read the latest post
				postRaw, err := ReadRaw(latestPost.OriginalPath)
				if err == nil {
					// Parse the post to HTML to display as preview
					// We need to parse it fully to get the HTML content
					// Re-using the MD parser config from above would be cleaner but let's instantiate for now
					mdPreview := goldmark.New(
						goldmark.WithExtensions(
							extension.GFM,
							&frontmatter.Extender{},
							mathjax.MathJax,
							templates.SidebarTag,
							templates.TopBanner,
						),
						goldmark.WithRendererOptions(html.WithUnsafe()),
					)
					
					var previewBuf bytes.Buffer
					ctxPreview := parser.NewContext()
					if err := mdPreview.Convert(postRaw, &previewBuf, parser.WithContext(ctxPreview)); err == nil {
						// Extract Metadata of the post to update Page Title?
						// User said: "the [title] be big text embedded in the black area."
						// So if we are previewing the latest post, the Home Page Title should probably match the Post Title on the Home Page?
						// Or should it say "Latest: Title"?
						// The main header h1 is where {{Title}} goes.
						// Getting metadata from preview context
						var postMeta templates.PageMetadata
						dPost := frontmatter.Get(ctxPreview)
						if dPost != nil {
							dPost.Decode(&postMeta)
						}
						
						// Fallback title logic
						if postMeta.Title == "" {
							base := filepath.Base(latestPost.OriginalPath)
							postMeta.Title = strings.TrimSuffix(base, filepath.Ext(base))
						}
						
						// Update the meta for the HOME PAGE render
						meta.Title = postMeta.Title // Use the post's title
						
						// Extract Preview (First Paragraph?)
						// "preview of the latest post, then read more"
						fullHTML := previewBuf.String()
						
						// Simple strategy: take everything up to the first </p> or limit chars?
						// Or explicit split?
						// Let's take the first paragraph.
						parts := strings.Split(fullHTML, "</p>")
						if len(parts) > 0 {
							previewHTML := parts[0] + "</p>"
							// Add Read More Link
							readMore := fmt.Sprintf(`<p><a href="%s">Read More...</a></p>`, latestPost.RoutePath)
							htmlContent = previewHTML + readMore
						} else {
							htmlContent = fullHTML
						}
					}
				}
			}
		}

		// [directory] Tag Expansion
		// User can put [directory] in any page (e.g. archive.md) to generate the list
		if strings.Contains(htmlContent, "[directory]") {
			// Generate Directory HTML (Grouped by Year)
			var dirHTML strings.Builder
			dirHTML.WriteString(`<div class="directory-list">`)
			
			// Group by year
			type YearGroup struct {
				Year  int
				Posts []templates.PostSnippet
			}
			var years []YearGroup
			
			for _, f := range AllContent {
				if f.IsMarkdown && f.RoutePath != "/" {
					name := filepath.Base(f.OriginalPath)
					title := strings.TrimSuffix(name, filepath.Ext(name))
					
					// Get Metadata for title? Ideally yes, but costly to parse all?
					// Let's rely on cached AllContent if we had metadata there, but we don't.
					// For now, filename title is fast.
					
					year := f.ModTime.Year()
					
					// Add to current year group or new one
					if len(years) == 0 || years[len(years)-1].Year != year {
						years = append(years, YearGroup{Year: year})
					}
					years[len(years)-1].Posts = append(years[len(years)-1].Posts, templates.PostSnippet{
						Title: title,
						URL:   f.RoutePath,
						Date:  f.ModTime,
					})
				}
			}
			
			for _, yg := range years {
				dirHTML.WriteString(fmt.Sprintf(`<h3>%d</h3><ul>`, yg.Year))
				for _, p := range yg.Posts {
					dirHTML.WriteString(fmt.Sprintf(`<li><a href="%s">%s</a> <span class="date">(%s)</span></li>`, 
						p.URL, p.Title, p.Date.Format("Jan 02")))
				}
				dirHTML.WriteString(`</ul>`)
			}
			dirHTML.WriteString(`</div>`)
			
			htmlContent = strings.ReplaceAll(htmlContent, "[directory]", dirHTML.String())
		}

			/**
			Add More Custom tags here. Add a customTag in tags.go as well, and the implementation in a template file.
			*/

			// ||| SideBar
			sidebarRegex := regexp.MustCompile(`(?s)<sidebar>.*?</sidebar>`)

			if sidebarRegex.MatchString(htmlContent) {
				// Render the dynamic content
				// Generate posts data
			    var posts []templates.PostSnippet
			    for _, f := range AllContent {
				    if f.IsMarkdown && f.RoutePath != "/" {
					    name := filepath.Base(f.OriginalPath)
					    title := strings.TrimSuffix(name, filepath.Ext(name))

					    posts = append(posts, templates.PostSnippet{
						    Title: title,
						    URL:   f.RoutePath,
						    Date:  f.ModTime,
					    })
				    }
			    }

				latestPostsHTML := templates.RenderLatestPosts(posts)
				dirLink := templates.RenderDirectoryLink()

				// Wrap in <div class="sidebar"> to match CSS
				replacement := fmt.Sprintf(`<div class="sidebar">%s%s</div>`, latestPostsHTML, dirLink)

				// Replace the placeholder tag with the actual content
				htmlContent = sidebarRegex.ReplaceAllString(htmlContent, replacement)
			}

			// _-_- TopBanner
			topbannerRegex := regexp.MustCompile(`(?s)<topbanner>.*?</topbanner>`)

			if topbannerRegex.MatchString(htmlContent) {
				htmlContent = topbannerRegex.ReplaceAllStringFunc(htmlContent, func(match string) string {
					// 'match' is "<topbanner>...content...</topbanner>"

					// Remove the known tags safely
					innerContent := strings.TrimPrefix(match, "<topbanner>")
					innerContent = strings.TrimSuffix(innerContent, "</topbanner>")

					// Pass to template
					bannerHTML := templates.RenderTopBanner(innerContent)

					// Return the formatted HTML
					return fmt.Sprintf(`<div class="topbanner">%s</div>`, bannerHTML)
				})
			}
		

		

			
			// [filename] Image Resolution
			// Matches [filename.ext] or [filename]
			// Avoids matches that look like markdown links [text](url) or existing tags
			// We'll use a slightly broad regex and rely on file existence check
			// Regex: \[([^\]<]+)\]
			// But careful not to match [text](url). 
			// Goldmark renders [text](url) as <a href="...">...</a>.
			// However, if the link is broken (reference link not defined)? It renders as [text].
			// So this strategy is acceptable per user request.
			
			imgRegex := regexp.MustCompile(`\[([^\]<]+)\]`)
			if imgRegex.MatchString(htmlContent) {
				htmlContent = imgRegex.ReplaceAllStringFunc(htmlContent, func(match string) string {
					// match is "[filename]"
					name := match[1 : len(match)-1] // Strip [ and ]
					
					// Resolve Image Path
					// Current directory of the document
					docDir := filepath.Dir(file.OriginalPath)
					
					// Root content directory
					rootDir := "content" 
					
					imagePath, err := resolveImage(name, docDir, rootDir)
					if err != nil {
						// Not found, return original text
						return match
					}
					
					// Calculate web path
					// imagePath is a filesystem path e.g. "content/entries/image.png"
					// We need "/entries/image.png"
					webPath := "/" + filepath.ToSlash(strings.TrimPrefix(imagePath, "content/"))
					
					// Check for styling metadata
					// Config key is the filename (e.g. "image.png")
					// Use the resolved filename for the key
					resolvedName := filepath.Base(imagePath)
					
					var styleAttr, otherAttrs string
					
					if configVal, ok := meta.Raw[resolvedName]; ok {
						// Config can be:
						// 1. A map (TOML: image = {width="...", ...})
						// 2. A string (TOML: image = "style=...") ? User example: image = "width: auto" ? No, user said: [a] style = "..." OR width = auto
						
						if configMap, ok := configVal.(map[string]interface{}); ok {
							for k, v := range configMap {
								val := fmt.Sprintf("%v", v)
								if k == "style" {
									styleAttr = fmt.Sprintf(` style="%s"`, val)
								} else {
									// height, width, etc.
									otherAttrs += fmt.Sprintf(` %s="%s"`, k, val)
								}
							}
						}
					}
					
					return fmt.Sprintf(`<img src="%s" alt="%s"%s%s>`, webPath, name, styleAttr, otherAttrs)
				})
			}

		// Build Full Page
		// Generate global posts data (Latest 5-10) for the sidebar
		// This logic is now handled within the sidebarRegex.MatchString block if a sidebar is present.
		// If no sidebar, we still need to generate the latest posts for the full page template.
		var globalPostsForTemplate []templates.PostSnippet
		for _, f := range AllContent {
			if f.IsMarkdown && f.RoutePath != "/" {
				name := filepath.Base(f.OriginalPath)
				title := strings.TrimSuffix(name, filepath.Ext(name))

				globalPostsForTemplate = append(globalPostsForTemplate, templates.PostSnippet{
					Title: title,
					URL:   f.RoutePath,
					Date:  f.ModTime,
				})
			}
			// Limit to 10 latest
			if len(globalPostsForTemplate) >= 10 {
				break
			}
		}
		
		// This `latestPostsHTML` is for the overall page template, not necessarily the sidebar content.
		// The sidebar content is generated and injected directly into `htmlContent`.
		// This `latestPostsHTML` is passed to `BuildFullPage` for the general layout.
		latestPostsHTML := templates.RenderLatestPosts(globalPostsForTemplate)
		latestPostsHTML += templates.RenderDirectoryLink()

		dataToWrite = templates.BuildFullPage([]byte(htmlContent), meta, latestPostsHTML)
	}

	// 4. Return Content Object or Path?
	// The function signature returns (string, error), which is the path to the cached file.
	// Since we are disabling cache, we should probably change the architecture to return ([]byte, error) 
	// or write to a temporary file if the rest of the app expects a file path.
	// The rest of the app (server.go) does http.ServeFile(w, r, cachePath).
	// So we MUST return a file path.
	
	// For now, we will continue to write to the "cache" location, but since we removed the read check,
	// it basically acts as a "render to temp file" on every request.
	// This satisfies the "disable caching" requirement (always fresh) while keeping the file-based serving architecture.
	
	tmpPath := cachePath + ".tmp"
	if err := os.WriteFile(tmpPath, dataToWrite, 0644); err != nil {
		return "", fmt.Errorf("failed to write cache: %w", err)
	}
	if err := os.Rename(tmpPath, cachePath); err != nil {
		return "", fmt.Errorf("failed to commit cache: %w", err)
	}

	return cachePath, nil
}

// loadExternalMetadata checks for .yml, .yaml, .toml files and loads them.
func loadExternalMetadata(markdownPath string, meta *templates.PageMetadata) error {
	basePath := strings.TrimSuffix(markdownPath, filepath.Ext(markdownPath))
	
	extensions := []string{".yml", ".yaml", ".toml"}
	
	for _, ext := range extensions {
		path := basePath + ext
		if _, err := os.Stat(path); err == nil {
			// Found a metadata file
			data, err := os.ReadFile(path)
			if err != nil {
				return fmt.Errorf("read file %s: %w", path, err)
			}
			
			// Decode into a map first to capture everything in Raw
			if meta.Raw == nil {
				meta.Raw = make(map[string]interface{})
			}
			
			if ext == ".toml" {
				if _, err := toml.Decode(string(data), &meta.Raw); err != nil {
					return fmt.Errorf("decode toml %s: %w", path, err)
				}
			} else {
				// yaml handles both .yml and .yaml
				if err := yaml.Unmarshal(data, &meta.Raw); err != nil {
					return fmt.Errorf("decode yaml %s: %w", path, err)
				}
			}
			
			// Map known fields to struct
			if v, ok := meta.Raw["title"]; ok {
				if s, ok := v.(string); ok { meta.Title = s }
			}
			if v, ok := meta.Raw["stylesheet"]; ok {
				if s, ok := v.(string); ok { meta.Stylesheet = s }
			}
			if v, ok := meta.Raw["style"]; ok {
				if s, ok := v.(string); ok { meta.Style = s }
			}
			
			// Stop after finding the first matching metadata file
			// Priority order is implicit in `extensions` slice (.yml > .yaml > .toml)
			return nil
		}
	}
	return nil
}

// resolveImage searches for the image file
func resolveImage(name, docDir, rootDir string) (string, error) {
	// 1. Check strict path in docDir
	path := filepath.Join(docDir, name)
	if !hasExt(name) {
		// Try extensions
		if found, err := findNewestWithExt(docDir, name); err == nil {
			return found, nil
		}
	} else {
		if _, err := os.Stat(path); err == nil {
			return path, nil
		}
	}
	
	// 2. Check strict path in rootDir
	path = filepath.Join(rootDir, name)
	if !hasExt(name) {
		if found, err := findNewestWithExt(rootDir, name); err == nil {
			return found, nil
		}
	} else {
		if _, err := os.Stat(path); err == nil {
			return path, nil
		}
	}
	
	return "", fmt.Errorf("not found")
}

func hasExt(name string) bool {
	return filepath.Ext(name) != ""
}

func findNewestWithExt(dir, baseName string) (string, error) {
	entries, err := os.ReadDir(dir)
	if err != nil {
		return "", err
	}
	
	var bestMatch string
	var newestTime time.Time
	
	targetBase := strings.ToLower(baseName)
	
	for _, e := range entries {
		if e.IsDir() {
			continue
		}
		
		name := e.Name()
		ext := filepath.Ext(name)
		if ext == "" {
			continue
		}
		
		nameNoExt := strings.TrimSuffix(name, ext)
		if strings.ToLower(nameNoExt) == targetBase {
			// Check if image extension
			info, err := e.Info()
			if err != nil {
				continue
			}
			
			if bestMatch == "" || info.ModTime().After(newestTime) {
				bestMatch = filepath.Join(dir, name)
				newestTime = info.ModTime()
			}
		}
	}
	
	if bestMatch != "" {
		return bestMatch, nil
	}
	return "", fmt.Errorf("not found")
}