all: Rework page store, add a dynacache, improve partial rebuilds, and some general spring cleaningdevelop2024

There are some breaking changes in this commit, see #11455.

Closes #11455
Closes #11549

This fixes a set of bugs (see issue list) and it is also paying some technical debt accumulated over the years. We now build with Staticcheck enabled in the CI build.

The performance should be about the same as before for regular sized Hugo sites, but it should perform and scale much better to larger data sets, as objects that uses lots of memory (e.g. rendered Markdown, big JSON files read into maps with transform.Unmarshal etc.) will now get automatically garbage collected if needed. Performance on partial rebuilds when running the server in fast render mode should be the same, but the change detection should be much more accurate.

A list of the notable new features:

* A new dependency tracker that covers (almost) all of Hugo's API and is used to do fine grained partial rebuilds when running the server.
* A new and simpler tree document store which allows fast lookups and prefix-walking in all dimensions (e.g. language) concurrently.
* You can now configure an upper memory limit allowing for much larger data sets and/or running on lower specced PCs.
We have lifted the "no resources in sub folders" restriction for branch bundles (e.g. sections).
Memory Limit
* Hugos will, by default, set aside a quarter of the total system memory, but you can set this via the OS environment variable HUGO_MEMORYLIMIT (in gigabytes). This is backed by a partitioned LRU cache used throughout Hugo. A cache that gets dynamically resized in low memory situations, allowing Go's Garbage Collector to free the memory.

New Dependency Tracker: Hugo has had a rule based coarse grained approach to server rebuilds that has worked mostly pretty well, but there have been some surprises (e.g. stale content). This is now revamped with a new dependency tracker that can quickly calculate the delta given a changed resource (e.g. a content file, template, JS file etc.). This handles transitive relations, e.g. $page -> js.Build -> JS import, or $page1.Content -> render hook -> site.GetPage -> $page2.Title, or $page1.Content -> shortcode -> partial -> site.RegularPages -> $page2.Content -> shortcode ..., and should also handle changes to aggregated values (e.g. site.Lastmod) effectively.

This covers all of Hugo's API with 2 known exceptions (a list that may not be fully exhaustive):

Changes to files loaded with template func os.ReadFile may not be handled correctly. We recommend loading resources with resources.Get
Changes to Hugo objects (e.g. Page) passed in the template context to lang.Translate may not be detected correctly. We recommend having simple i18n templates without too much data context passed in other than simple types such as strings and numbers.
Note that the cachebuster configuration (when A changes then rebuild B) works well with the above, but we recommend that you revise that configuration, as it in most situations should not be needed. One example where it is still needed is with TailwindCSS and using changes to hugo_stats.json to trigger new CSS rebuilds.

Document Store: Previously, a little simplified, we split the document store (where we store pages and resources) in a tree per language. This worked pretty well, but the structure made some operations harder than they needed to be. We have now restructured it into one Radix tree for all languages. Internally the language is considered to be a dimension of that tree, and the tree can be viewed in all dimensions concurrently. This makes some operations re. language simpler (e.g. finding translations is just a slice range), but the idea is that it should also be relatively inexpensive to add more dimensions if needed (e.g. role).

Fixes #10169
Fixes #10364
Fixes #10482
Fixes #10630
Fixes #10656
Fixes #10694
Fixes #10918
Fixes #11262
Fixes #11439
Fixes #11453
Fixes #11457
Fixes #11466
Fixes #11540
Fixes #11551
Fixes #11556
Fixes #11654
Fixes #11661
Fixes #11663
Fixes #11664
Fixes #11669
Fixes #11671
Fixes #11807
Fixes #11808
Fixes #11809
Fixes #11815
Fixes #11840
Fixes #11853
Fixes #11860
Fixes #11883
Fixes #11904
Fixes #7388
Fixes #7425
Fixes #7436
Fixes #7544
Fixes #7882
Fixes #7960
Fixes #8255
Fixes #8307
Fixes #8863
Fixes #8927
Fixes #9192
Fixes #9324

1 files changed, 519 insertions, 94 deletions

diff --git a/hugolib/hugo_sites_build.go b/hugolib/hugo_sites_build.go
index a2b87b5e7..257949334 100644
--- a/hugolib/hugo_sites_build.go
+++ b/hugolib/hugo_sites_build.go
@@ -1,4 +1,4 @@
-// Copyright 2019 The Hugo Authors. All rights reserved.
+// Copyright 2024 The Hugo Authors. All rights reserved.
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
@@ -19,38 +19,47 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
+	"os"
+	"path"
 	"path/filepath"
 	"strings"
 	"time"
 
 	"github.com/bep/logg"
+	"github.com/gohugoio/hugo/cache/dynacache"
+	"github.com/gohugoio/hugo/deps"
 	"github.com/gohugoio/hugo/hugofs/files"
+	"github.com/gohugoio/hugo/hugofs/glob"
+	"github.com/gohugoio/hugo/identity"
+	"github.com/gohugoio/hugo/output"
 	"github.com/gohugoio/hugo/publisher"
+	"github.com/gohugoio/hugo/source"
 	"github.com/gohugoio/hugo/tpl"
 
 	"github.com/gohugoio/hugo/hugofs"
 
+	"github.com/gohugoio/hugo/common/herrors"
 	"github.com/gohugoio/hugo/common/loggers"
 	"github.com/gohugoio/hugo/common/para"
+	"github.com/gohugoio/hugo/common/paths"
 	"github.com/gohugoio/hugo/config"
+	"github.com/gohugoio/hugo/resources/page"
+	"github.com/gohugoio/hugo/resources/page/siteidentities"
 	"github.com/gohugoio/hugo/resources/postpub"
 
 	"github.com/spf13/afero"
 
-	"github.com/gohugoio/hugo/output"
-
 	"github.com/fsnotify/fsnotify"
 )
 
 // Build builds all sites. If filesystem events are provided,
 // this is considered to be a potential partial rebuild.
 func (h *HugoSites) Build(config BuildCfg, events ...fsnotify.Event) error {
-	if h == nil {
-		return errors.New("cannot build nil *HugoSites")
-	}
+	infol := h.Log.InfoCommand("build")
+	defer loggers.TimeTrackf(infol, time.Now(), nil, "")
 
 	if h.Deps == nil {
-		return errors.New("cannot build nil *Deps")
+		panic("must have deps")
 	}
 
 	if !config.NoBuildLock {
@@ -67,8 +76,6 @@ func (h *HugoSites) Build(config BuildCfg, events ...fsnotify.Event) error {
 		}
 	}()
 
-	infol := h.Log.InfoCommand("build")
-
 	errCollector := h.StartErrorCollector()
 	errs := make(chan error)
 
@@ -91,14 +98,16 @@ func (h *HugoSites) Build(config BuildCfg, events ...fsnotify.Event) error {
 		h.Metrics.Reset()
 	}
 
-	h.testCounters = config.testCounters
+	h.buildCounters = config.testCounters
+	if h.buildCounters == nil {
+		h.buildCounters = &buildCounters{}
+	}
 
 	// Need a pointer as this may be modified.
 	conf := &config
-
 	if conf.whatChanged == nil {
 		// Assume everything has changed
-		conf.whatChanged = &whatChanged{source: true}
+		conf.whatChanged = &whatChanged{contentChanged: true}
 	}
 
 	var prepareErr error
@@ -124,11 +133,13 @@ func (h *HugoSites) Build(config BuildCfg, events ...fsnotify.Event) error {
 				return nil
 			}
 
-			if err := h.process(infol, conf, init, events...); err != nil {
+			ctx := context.Background()
+
+			if err := h.process(ctx, infol, conf, init, events...); err != nil {
 				return fmt.Errorf("process: %w", err)
 			}
 
-			if err := h.assemble(infol, conf); err != nil {
+			if err := h.assemble(ctx, infol, conf); err != nil {
 				return fmt.Errorf("assemble: %w", err)
 			}
 
@@ -190,16 +201,17 @@ func (h *HugoSites) initSites(config *BuildCfg) error {
 }
 
 func (h *HugoSites) initRebuild(config *BuildCfg) error {
-	if config.ResetState {
-		return errors.New("rebuild does not support 'ResetState'")
-	}
-
 	if !h.Configs.Base.Internal.Watch {
 		return errors.New("rebuild called when not in watch mode")
 	}
 
+	h.pageTrees.treePagesResources.WalkPrefixRaw("", func(key string, n contentNodeI) bool {
+		n.resetBuildState()
+		return false
+	})
+
 	for _, s := range h.Sites {
-		s.resetBuildState(config.whatChanged.source)
+		s.resetBuildState(config.whatChanged.contentChanged)
 	}
 
 	h.reset(config)
@@ -208,70 +220,102 @@ func (h *HugoSites) initRebuild(config *BuildCfg) error {
 	return nil
 }
 
-func (h *HugoSites) process(l logg.LevelLogger, config *BuildCfg, init func(config *BuildCfg) error, events ...fsnotify.Event) error {
-	defer h.timeTrack(l, time.Now(), "process")
-
-	// We should probably refactor the Site and pull up most of the logic from there to here,
-	// but that seems like a daunting task.
-	// So for now, if there are more than one site (language),
-	// we pre-process the first one, then configure all the sites based on that.
-
-	firstSite := h.Sites[0]
+// process prepares the Sites' sources for a full or partial rebuild.
+// This will also parse the source and create all the Page objects.
+func (h *HugoSites) process(ctx context.Context, l logg.LevelLogger, config *BuildCfg, init func(config *BuildCfg) error, events ...fsnotify.Event) error {
+	l = l.WithField("step", "process")
+	defer loggers.TimeTrackf(l, time.Now(), nil, "")
 
 	if len(events) > 0 {
 		// This is a rebuild
-		return firstSite.processPartial(config, init, events)
+		return h.processPartial(ctx, l, config, init, events)
 	}
-
-	return firstSite.process(*config)
+	return h.processFull(ctx, l, *config)
 }
 
-func (h *HugoSites) assemble(l logg.LevelLogger, bcfg *BuildCfg) error {
-	defer h.timeTrack(l, time.Now(), "assemble")
+// assemble creates missing sections, applies agregate values (e.g. dates, cascading params),
+// removes disabled pages etc.
+func (h *HugoSites) assemble(ctx context.Context, l logg.LevelLogger, bcfg *BuildCfg) error {
+	l = l.WithField("step", "assemble")
+	defer loggers.TimeTrackf(l, time.Now(), nil, "")
 
-	if !bcfg.whatChanged.source {
+	if !bcfg.whatChanged.contentChanged {
 		return nil
 	}
 
-	if err := h.getContentMaps().AssemblePages(); err != nil {
-		return err
+	h.translationKeyPages.Reset()
+	assemblers := make([]*sitePagesAssembler, len(h.Sites))
+	// Changes detected during assembly (e.g. aggregate date changes)
+	assembleChanges := &whatChanged{
+		identitySet: make(map[identity.Identity]bool),
+	}
+	for i, s := range h.Sites {
+		assemblers[i] = &sitePagesAssembler{
+			Site:            s,
+			watching:        s.watching(),
+			incomingChanges: bcfg.whatChanged,
+			assembleChanges: assembleChanges,
+			ctx:             ctx,
+		}
 	}
 
-	if err := h.createPageCollections(); err != nil {
+	g, _ := h.workersSite.Start(ctx)
+	for _, s := range assemblers {
+		s := s
+		g.Run(func() error {
+			return s.assemblePagesStep1(ctx)
+		})
+	}
+	if err := g.Wait(); err != nil {
 		return err
 	}
 
-	return nil
-}
+	changes := assembleChanges.Changes()
 
-func (h *HugoSites) timeTrack(l logg.LevelLogger, start time.Time, name string) {
-	elapsed := time.Since(start)
-	l.WithField("step", name).WithField("duration", elapsed).Logf("running")
+	// Changes from the assemble step (e.g. lastMod, cascase) needs a re-calculation
+	// of what needs to be re-built.
+	if len(changes) > 0 {
+		if err := h.resolveAndClearStateForIdentities(ctx, l, nil, changes); err != nil {
+			return err
+		}
+	}
+	h.renderFormats = output.Formats{}
+	for _, s := range h.Sites {
+		s.s.initRenderFormats()
+		h.renderFormats = append(h.renderFormats, s.renderFormats...)
+	}
+
+	for _, s := range assemblers {
+		if err := s.assemblePagesStep2(); err != nil {
+			return err
+		}
+	}
+
+	h.renderFormats = output.Formats{}
+	for _, s := range h.Sites {
+		h.renderFormats = append(h.renderFormats, s.renderFormats...)
+	}
+
+	return nil
 }
 
+// render renders the sites.
 func (h *HugoSites) render(l logg.LevelLogger, config *BuildCfg) error {
-	defer h.timeTrack(l, time.Now(), "render")
+	l = l.WithField("step", "render")
+	start := time.Now()
+	defer func() {
+		loggers.TimeTrackf(l, start, h.buildCounters.loggFields(), "")
+	}()
+
 	if _, err := h.init.layouts.Do(context.Background()); err != nil {
 		return err
 	}
 
 	siteRenderContext := &siteRenderContext{cfg: config, multihost: h.Configs.IsMultihost}
 
-	if !config.PartialReRender {
-		h.renderFormats = output.Formats{}
-		h.withSite(func(s *Site) error {
-			s.initRenderFormats()
-			return nil
-		})
-
-		for _, s := range h.Sites {
-			h.renderFormats = append(h.renderFormats, s.renderFormats...)
-		}
-	}
-
 	i := 0
-
 	for _, s := range h.Sites {
+		siteRenderContext.languageIdx = s.languagei
 		h.currentSite = s
 		for siteOutIdx, renderFormat := range s.renderFormats {
 			siteRenderContext.outIdx = siteOutIdx
@@ -293,6 +337,12 @@ func (h *HugoSites) render(l logg.LevelLogger, config *BuildCfg) error {
 					}
 				}
 				if !config.SkipRender {
+					ll := l.WithField("substep", "pages").
+						WithField("site", s.language.Lang).
+						WithField("outputFormat", renderFormat.Name)
+
+					start := time.Now()
+
 					if config.PartialReRender {
 						if err := s.renderPages(siteRenderContext); err != nil {
 							return err
@@ -302,24 +352,16 @@ func (h *HugoSites) render(l logg.LevelLogger, config *BuildCfg) error {
 							return err
 						}
 					}
+					loggers.TimeTrackf(ll, start, nil, "")
 				}
 			}
-
-		}
-	}
-
-	if !config.SkipRender {
-		if err := h.renderCrossSitesSitemap(); err != nil {
-			return err
-		}
-		if err := h.renderCrossSitesRobotsTXT(); err != nil {
-			return err
 		}
 	}
 
 	return nil
 }
 
+// / postRenderOnce runs some post processing that only needs to be done once, e.g. printing of unused templates.
 func (h *HugoSites) postRenderOnce() error {
 	h.postRenderInit.Do(func() {
 		conf := h.Configs.Base
@@ -347,8 +389,10 @@ func (h *HugoSites) postRenderOnce() error {
 	return nil
 }
 
+// postProcess runs the post processors, e.g. writing the hugo_stats.json file.
 func (h *HugoSites) postProcess(l logg.LevelLogger) error {
-	defer h.timeTrack(l, time.Now(), "postProcess")
+	l = l.WithField("step", "postProcess")
+	defer loggers.TimeTrackf(l, time.Now(), nil, "")
 
 	// Make sure to write any build stats to disk first so it's available
 	// to the post processors.
@@ -360,35 +404,40 @@ func (h *HugoSites) postProcess(l logg.LevelLogger) error {
 	// imports that resolves to the project or a module.
 	// Write a jsconfig.json file to the project's /asset directory
 	// to help JS IntelliSense in VS Code etc.
-	if !h.ResourceSpec.BuildConfig().NoJSConfigInAssets && h.BaseFs.Assets.Dirs != nil {
-		fi, err := h.BaseFs.Assets.Fs.Stat("")
-		if err != nil {
-			h.Log.Warnf("Failed to resolve jsconfig.json dir: %s", err)
-		} else {
+	if !h.ResourceSpec.BuildConfig().NoJSConfigInAssets {
+		handleJSConfig := func(fi os.FileInfo) {
 			m := fi.(hugofs.FileMetaInfo).Meta()
-			assetsDir := m.SourceRoot
-			if strings.HasPrefix(assetsDir, h.Configs.LoadingInfo.BaseConfig.WorkingDir) {
-				if jsConfig := h.ResourceSpec.JSConfigBuilder.Build(assetsDir); jsConfig != nil {
+			if !m.IsProject {
+				return
+			}
 
-					b, err := json.MarshalIndent(jsConfig, "", " ")
-					if err != nil {
-						h.Log.Warnf("Failed to create jsconfig.json: %s", err)
-					} else {
-						filename := filepath.Join(assetsDir, "jsconfig.json")
-						if h.Configs.Base.Internal.Running {
-							h.skipRebuildForFilenamesMu.Lock()
-							h.skipRebuildForFilenames[filename] = true
-							h.skipRebuildForFilenamesMu.Unlock()
-						}
-						// Make sure it's  written to the OS fs as this is used by
-						// editors.
-						if err := afero.WriteFile(hugofs.Os, filename, b, 0o666); err != nil {
-							h.Log.Warnf("Failed to write jsconfig.json: %s", err)
-						}
+			if jsConfig := h.ResourceSpec.JSConfigBuilder.Build(m.SourceRoot); jsConfig != nil {
+				b, err := json.MarshalIndent(jsConfig, "", " ")
+				if err != nil {
+					h.Log.Warnf("Failed to create jsconfig.json: %s", err)
+				} else {
+					filename := filepath.Join(m.SourceRoot, "jsconfig.json")
+					if h.Configs.Base.Internal.Running {
+						h.skipRebuildForFilenamesMu.Lock()
+						h.skipRebuildForFilenames[filename] = true
+						h.skipRebuildForFilenamesMu.Unlock()
+					}
+					// Make sure it's  written to the OS fs as this is used by
+					// editors.
+					if err := afero.WriteFile(hugofs.Os, filename, b, 0o666); err != nil {
+						h.Log.Warnf("Failed to write jsconfig.json: %s", err)
 					}
 				}
 			}
+		}
 
+		fi, err := h.BaseFs.Assets.Fs.Stat("")
+		if err != nil {
+			if !herrors.IsNotExist(err) {
+				h.Log.Warnf("Failed to resolve jsconfig.json dir: %s", err)
+			}
+		} else {
+			handleJSConfig(fi)
 		}
 	}
 
@@ -466,10 +515,6 @@ func (h *HugoSites) postProcess(l logg.LevelLogger) error {
 	return g.Wait()
 }
 
-type publishStats struct {
-	CSSClasses string `json:"cssClasses"`
-}
-
 func (h *HugoSites) writeBuildStats() error {
 	if h.ResourceSpec == nil {
 		panic("h.ResourceSpec is nil")
@@ -523,3 +568,383 @@ func (h *HugoSites) writeBuildStats() error {
 
 	return nil
 }
+
+type pathChange struct {
+	// The path to the changed file.
+	p *paths.Path
+
+	// If true, this is a delete operation (a delete or a rename).
+	delete bool
+
+	// If true, this is a directory.
+	isDir bool
+}
+
+// processPartial prepares the Sites' sources for a partial rebuild.
+func (h *HugoSites) processPartial(ctx context.Context, l logg.LevelLogger, config *BuildCfg, init func(config *BuildCfg) error, events []fsnotify.Event) error {
+	h.Log.Trace(logg.StringFunc(func() string {
+		var sb strings.Builder
+		sb.WriteString("File events:\n")
+		for _, ev := range events {
+			sb.WriteString(ev.String())
+			sb.WriteString("\n")
+		}
+		return sb.String()
+	}))
+
+	events = h.fileEventsFilter(events)
+	events = h.fileEventsTranslate(events)
+
+	logger := h.Log
+
+	var (
+		tmplAdded      bool
+		tmplChanged    bool
+		i18nChanged    bool
+		contentChanged bool
+	)
+
+	changedPaths := struct {
+		changedFiles []*paths.Path
+		changedDirs  []*paths.Path
+		deleted      []*paths.Path
+	}{}
+
+	removeDuplicatePaths := func(ps []*paths.Path) []*paths.Path {
+		seen := make(map[string]bool)
+		var filtered []*paths.Path
+		for _, p := range ps {
+			if !seen[p.Path()] {
+				seen[p.Path()] = true
+				filtered = append(filtered, p)
+			}
+		}
+		return filtered
+	}
+
+	var (
+		cacheBusters      []func(string) bool
+		deletedDirs       []string
+		addedContentPaths []*paths.Path
+	)
+
+	for _, ev := range events {
+		removed := false
+		added := false
+
+		if ev.Op&fsnotify.Remove == fsnotify.Remove {
+			removed = true
+		}
+
+		fi, statErr := h.Fs.Source.Stat(ev.Name)
+
+		// Some editors (Vim) sometimes issue only a Rename operation when writing an existing file
+		// Sometimes a rename operation means that file has been renamed other times it means
+		// it's been updated.
+		if ev.Op.Has(fsnotify.Rename) {
+			// If the file is still on disk, it's only been updated, if it's not, it's been moved
+			if statErr != nil {
+				removed = true
+			}
+		}
+		if ev.Op.Has(fsnotify.Create) {
+			added = true
+		}
+
+		isChangedDir := statErr == nil && fi.IsDir()
+
+		cpss := h.BaseFs.ResolvePaths(ev.Name, !removed)
+		pss := make([]*paths.Path, len(cpss))
+		for i, cps := range cpss {
+			p := cps.Path
+			if removed && !paths.HasExt(p) {
+				// Assume this is a renamed/removed directory.
+				// For deletes, we walk up the tree to find the container (e.g. branch bundle),
+				// so we will catch this even if it is a file without extension.
+				// This avoids us walking up to the home page bundle for the common case
+				// of renaming root sections.
+				p = p + "/_index.md"
+				deletedDirs = append(deletedDirs, cps.Path)
+			}
+
+			pss[i] = h.Configs.ContentPathParser.Parse(cps.Component, p)
+			if added && !isChangedDir && cps.Component == files.ComponentFolderContent {
+				addedContentPaths = append(addedContentPaths, pss[i])
+			}
+
+			// Compile cache buster.
+			np := glob.NormalizePath(path.Join(cps.Component, cps.Path))
+			g, err := h.ResourceSpec.BuildConfig().MatchCacheBuster(h.Log, np)
+			if err == nil && g != nil {
+				cacheBusters = append(cacheBusters, g)
+			}
+		}
+
+		if removed {
+			changedPaths.deleted = append(changedPaths.deleted, pss...)
+		} else if isChangedDir {
+			changedPaths.changedDirs = append(changedPaths.changedDirs, pss...)
+		} else {
+			changedPaths.changedFiles = append(changedPaths.changedFiles, pss...)
+		}
+	}
+
+	var (
+		addedOrChangedContent []pathChange
+		changes               []identity.Identity
+	)
+
+	// Find the most specific identity possible.
+	handleChange := func(pathInfo *paths.Path, delete, isDir bool) {
+		switch pathInfo.Component() {
+		case files.ComponentFolderContent:
+			logger.Println("Source changed", pathInfo.Path())
+			base := pathInfo.Base()
+
+			if ids := h.pageTrees.collectIdentities(base); len(ids) > 0 {
+				changes = append(changes, ids...)
+			}
+
+			contentChanged = true
+
+			if config.RecentlyVisited != nil {
+				// Fast render mode. Adding them to the visited queue
+				// avoids rerendering them on navigation.
+				for _, id := range changes {
+					if p, ok := id.(page.Page); ok {
+						config.RecentlyVisited.Add(p.RelPermalink())
+					}
+				}
+			}
+
+			h.pageTrees.treeTaxonomyEntries.DeletePrefix("")
+
+			if delete {
+				_, ok := h.pageTrees.treePages.LongestPrefixAll(pathInfo.Base())
+				if ok {
+					h.pageTrees.treePages.DeleteAll(pathInfo.Base())
+					if pathInfo.IsBundle() {
+						// Assume directory removed.
+						h.pageTrees.treePages.DeletePrefixAll(pathInfo.Base() + "/")
+						h.pageTrees.resourceTrees.DeletePrefixAll(pathInfo.Base() + "/")
+					}
+				} else {
+					h.pageTrees.resourceTrees.DeleteAll(pathInfo.Base())
+				}
+			}
+
+			addedOrChangedContent = append(addedOrChangedContent, pathChange{p: pathInfo, delete: delete, isDir: isDir})
+
+		case files.ComponentFolderLayouts:
+			tmplChanged = true
+			h.init.layouts.Reset()
+			templatePath := pathInfo.TrimLeadingSlash().PathNoLang()
+			if !h.Tmpl().HasTemplate(templatePath) {
+				tmplAdded = true
+			}
+
+			if tmplAdded {
+				logger.Println("Template added", pathInfo.Path())
+				// A new template may require a more coarse grained build.
+				base := pathInfo.Base()
+				if strings.Contains(base, "_markup") {
+					// It's hard to determine the exact change set of this,
+					// so be very coarse grained.
+					changes = append(changes, identity.GenghisKhan)
+				}
+				if strings.Contains(base, "shortcodes") {
+					changes = append(changes, identity.NewGlobIdentity(fmt.Sprintf("shortcodes/%s*", pathInfo.BaseNameNoIdentifier())))
+				} else {
+					changes = append(changes, pathInfo)
+				}
+			} else {
+				logger.Println("Template changed", pathInfo.Path())
+				if templ, found := h.Tmpl().GetIdentity(templatePath); found {
+					changes = append(changes, templ)
+				} else {
+					changes = append(changes, pathInfo)
+				}
+			}
+		case files.ComponentFolderAssets:
+			logger.Println("Asset changed", pathInfo.Path())
+			r, _ := h.ResourceSpec.ResourceCache.Get(context.Background(), dynacache.CleanKey(pathInfo.Base()))
+			var hasID bool
+			identity.WalkIdentitiesShallow(r, func(level int, rid identity.Identity) bool {
+				hasID = true
+				changes = append(changes, rid)
+				return false
+			})
+			if !hasID {
+				changes = append(changes, pathInfo)
+			}
+		case files.ComponentFolderData:
+			logger.Println("Data changed", pathInfo.Path())
+
+			// This should cover all usage of site.Data.
+			// Currently very coarse grained.
+			changes = append(changes, siteidentities.Data)
+			h.init.data.Reset()
+		case files.ComponentFolderI18n:
+			logger.Println("i18n changed", pathInfo.Path())
+			i18nChanged = true
+			changes = append(changes, pathInfo)
+		default:
+			panic(fmt.Sprintf("unknown component: %q", pathInfo.Component()))
+		}
+	}
+
+	changedPaths.deleted = removeDuplicatePaths(changedPaths.deleted)
+	changedPaths.changedFiles = removeDuplicatePaths(changedPaths.changedFiles)
+
+	h.Log.Trace(logg.StringFunc(func() string {
+		var sb strings.Builder
+		sb.WriteString("Resolved paths:\n")
+		sb.WriteString("Deleted:\n")
+		for _, p := range changedPaths.deleted {
+			sb.WriteString("path: " + p.Path())
+			sb.WriteString("\n")
+		}
+		sb.WriteString("Changed:\n")
+		for _, p := range changedPaths.changedFiles {
+			sb.WriteString("path: " + p.Path())
+			sb.WriteString("\n")
+		}
+		return sb.String()
+	}))
+
+	for _, deletedDir := range deletedDirs {
+		prefix := deletedDir + "/"
+		predicate := func(id identity.Identity) bool {
+			// This will effectively reset all pages below this dir.
+			return strings.HasPrefix(paths.AddLeadingSlash(id.IdentifierBase()), prefix)
+		}
+		// Test in both directions.
+		changes = append(changes, identity.NewPredicateIdentity(
+			// Is dependent.
+			predicate,
+			// Is dependency.
+			predicate,
+		),
+		)
+	}
+
+	if len(addedContentPaths) > 0 {
+		// These content files are new and not in use anywhere.
+		// To make sure that these gets listed in any site.RegularPages ranges or similar
+		// we could invalidate everything, but first try to collect a sample set
+		// from the surrounding pages.
+		var surroundingIDs []identity.Identity
+		for _, p := range addedContentPaths {
+			if ids := h.pageTrees.collectIdentitiesSurrounding(p.Base(), 10); len(ids) > 0 {
+				surroundingIDs = append(surroundingIDs, ids...)
+			}
+		}
+
+		if len(surroundingIDs) > 0 {
+			changes = append(changes, surroundingIDs...)
+		} else {
+			// No surrounding pages found, so invalidate everything.
+			changes = append(changes, identity.GenghisKhan)
+		}
+	}
+
+	for _, deleted := range changedPaths.deleted {
+		handleChange(deleted, true, false)
+	}
+
+	for _, id := range changedPaths.changedFiles {
+		handleChange(id, false, false)
+	}
+
+	for _, id := range changedPaths.changedDirs {
+		handleChange(id, false, true)
+	}
+
+	resourceFiles := h.fileEventsContentPaths(addedOrChangedContent)
+
+	changed := &whatChanged{
+		contentChanged: contentChanged,
+		identitySet:    make(identity.Identities),
+	}
+	changed.Add(changes...)
+
+	config.whatChanged = changed
+
+	if err := init(config); err != nil {
+		return err
+	}
+
+	var cacheBusterOr func(string) bool
+	if len(cacheBusters) > 0 {
+		cacheBusterOr = func(s string) bool {
+			for _, cb := range cacheBusters {
+				if cb(s) {
+					return true
+				}
+			}
+			return false
+		}
+	}
+
+	// Removes duplicates.
+	changes = changed.identitySet.AsSlice()
+
+	if err := h.resolveAndClearStateForIdentities(ctx, l, cacheBusterOr, changes); err != nil {
+		return err
+	}
+
+	if tmplChanged || i18nChanged {
+		if err := loggers.TimeTrackfn(func() (logg.LevelLogger, error) {
+			// TODO(bep) this could probably be optimized to somehow
+			// only load the changed templates and its dependencies, but that is non-trivial.
+			ll := l.WithField("substep", "rebuild templates")
+			var prototype *deps.Deps
+			for i, s := range h.Sites {
+				if err := s.Deps.Compile(prototype); err != nil {
+					return ll, err
+				}
+				if i == 0 {
+					prototype = s.Deps
+				}
+			}
+			return ll, nil
+		}); err != nil {
+			return err
+		}
+	}
+
+	if resourceFiles != nil {
+		if err := h.processFiles(ctx, l, *config, resourceFiles...); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+func (h *HugoSites) processFull(ctx context.Context, l logg.LevelLogger, config BuildCfg) (err error) {
+	if err = h.processFiles(ctx, l, config); err != nil {
+		err = fmt.Errorf("readAndProcessContent: %w", err)
+		return
+	}
+	return err
+}
+
+func (s *HugoSites) processFiles(ctx context.Context, l logg.LevelLogger, buildConfig BuildCfg, filenames ...pathChange) error {
+	if s.Deps == nil {
+		panic("nil deps on site")
+	}
+
+	sourceSpec := source.NewSourceSpec(s.PathSpec, buildConfig.ContentInclusionFilter, s.BaseFs.Content.Fs)
+
+	// For inserts, we can pick an arbitrary pageMap.
+	pageMap := s.Sites[0].pageMap
+
+	c := newPagesCollector(ctx, s.h, sourceSpec, s.Log, l, pageMap, filenames)
+
+	if err := c.Collect(); err != nil {
+		return err
+	}
+
+	return nil
+}