[chore] media and emoji refactoring (#3000)

* start updating media manager interface ready for storing attachments / emoji right away

* store emoji and media as uncached immediately, then (re-)cache on Processing{}.Load()

* remove now unused media workers

* fix tests and issues

* fix another test!

* fix emoji activitypub uri setting behaviour, fix remainder of test compilation issues

* fix more tests

* fix (most of) remaining tests, add debouncing to repeatedly failing media / emojis

* whoops, rebase issue

* remove kim's whacky experiments

* do some reshuffling, ensure emoji uri gets set

* ensure marked as not cached on cleanup

* tweaks to media / emoji processing to handle context canceled better

* ensure newly fetched emojis actually get set in returned slice

* use different varnames to be a bit more obvious

* move emoji refresh rate limiting to dereferencer

* add exported dereferencer functions for remote media, use these for recaching in processor

* add check for nil attachment in updateAttachment()

* remove unused emoji and media fields + columns

* see previous commit

* fix old migrations expecting image_updated_at to exists (from copies of old models)

* remove freshness checking code (seems to be broken...)

* fix error arg causing nil ptr exception

* finish documentating functions with comments, slight tweaks to media / emoji deref error logic

* remove some extra unneeded boolean checking

* finish writing documentation (code comments) for exported media manager methods

* undo changes to migration snapshot gtsmodels, updated failing migration to have its own snapshot

* move doesColumnExist() to util.go in migrations package

internal/processing/media/getfile.go

@@ -19,14 +19,14 @@ package media
 
 import (
 	"context"
+	"errors"
 	"fmt"
-	"io"
 	"net/url"
 	"strings"
 	"time"
 
 	apimodel "github.com/superseriousbusiness/gotosocial/internal/api/model"
-	"github.com/superseriousbusiness/gotosocial/internal/gtscontext"
+	"github.com/superseriousbusiness/gotosocial/internal/db"
 	"github.com/superseriousbusiness/gotosocial/internal/gtserror"
 	"github.com/superseriousbusiness/gotosocial/internal/gtsmodel"
 	"github.com/superseriousbusiness/gotosocial/internal/media"
@@ -38,7 +38,7 @@ import (
 // to the caller via an io.reader embedded in *apimodel.Content.
 func (p *Processor) GetFile(
 	ctx context.Context,
-	requestingAccount *gtsmodel.Account,
+	requester *gtsmodel.Account,
 	form *apimodel.GetContentRequestForm,
 ) (*apimodel.Content, gtserror.WithCode) {
 	// parse the form fields
@@ -69,13 +69,13 @@ func (p *Processor) GetFile(
 	}
 
 	// make sure the requesting account and the media account don't block each other
-	if requestingAccount != nil {
-		blocked, err := p.state.DB.IsEitherBlocked(ctx, requestingAccount.ID, owningAccountID)
+	if requester != nil {
+		blocked, err := p.state.DB.IsEitherBlocked(ctx, requester.ID, owningAccountID)
 		if err != nil {
-			return nil, gtserror.NewErrorNotFound(fmt.Errorf("block status could not be established between accounts %s and %s: %s", owningAccountID, requestingAccount.ID, err))
+			return nil, gtserror.NewErrorNotFound(fmt.Errorf("block status could not be established between accounts %s and %s: %s", owningAccountID, requester.ID, err))
 		}
 		if blocked {
-			return nil, gtserror.NewErrorNotFound(fmt.Errorf("block exists between accounts %s and %s", owningAccountID, requestingAccount.ID))
+			return nil, gtserror.NewErrorNotFound(fmt.Errorf("block exists between accounts %s and %s", owningAccountID, requester.ID))
 		}
 	}
 
@@ -83,17 +83,254 @@ func (p *Processor) GetFile(
 	// so we need to take different steps depending on the media type being requested
 	switch mediaType {
 	case media.TypeEmoji:
-		return p.getEmojiContent(ctx, wantedMediaID, owningAccountID, mediaSize)
+		return p.getEmojiContent(ctx,
+			owningAccountID,
+			wantedMediaID,
+			mediaSize,
+		)
 	case media.TypeAttachment, media.TypeHeader, media.TypeAvatar:
-		return p.getAttachmentContent(ctx, requestingAccount, wantedMediaID, owningAccountID, mediaSize)
+		return p.getAttachmentContent(ctx,
+			requester,
+			owningAccountID,
+			wantedMediaID,
+			mediaSize,
+		)
 	default:
 		return nil, gtserror.NewErrorNotFound(fmt.Errorf("media type %s not recognized", mediaType))
 	}
 }
 
-/*
-	UTIL FUNCTIONS
-*/
+func (p *Processor) getAttachmentContent(
+	ctx context.Context,
+	requester *gtsmodel.Account,
+	ownerID string,
+	mediaID string,
+	sizeStr media.Size,
+) (
+	*apimodel.Content,
+	gtserror.WithCode,
+) {
+	// Search for media with given ID in the database.
+	attach, err := p.state.DB.GetAttachmentByID(ctx, mediaID)
+	if err != nil && !errors.Is(err, db.ErrNoEntries) {
+		err := gtserror.Newf("error fetching media from database: %w", err)
+		return nil, gtserror.NewErrorInternalError(err)
+	}
+
+	if attach == nil {
+		const text = "media not found"
+		return nil, gtserror.NewErrorNotFound(errors.New(text), text)
+	}
+
+	// Ensure the 'owner' owns media.
+	if attach.AccountID != ownerID {
+		const text = "media was not owned by passed account id"
+		return nil, gtserror.NewErrorNotFound(errors.New(text) /* no help text! */)
+	}
+
+	var remoteURL *url.URL
+	if attach.RemoteURL != "" {
+
+		// Parse media remote URL to valid URL object.
+		remoteURL, err = url.Parse(attach.RemoteURL)
+		if err != nil {
+			err := gtserror.Newf("invalid media remote url %s: %w", attach.RemoteURL, err)
+			return nil, gtserror.NewErrorInternalError(err)
+		}
+	}
+
+	// Uknown file types indicate no *locally*
+	// stored data we can serve. Handle separately.
+	if attach.Type == gtsmodel.FileTypeUnknown {
+		if remoteURL == nil {
+			err := gtserror.Newf("missing remote url for unknown type media %s: %w", attach.ID, err)
+			return nil, gtserror.NewErrorInternalError(err)
+		}
+
+		// If this is an "Unknown" file type, ie., one we
+		// tried to process and couldn't, or one we refused
+		// to process because it wasn't supported, then we
+		// can skip a lot of steps here by simply forwarding
+		// the request to the remote URL.
+		url := &storage.PresignedURL{
+			URL: remoteURL,
+
+			// We might manage to cache the media
+			// at some point, so set a low-ish expiry.
+			Expiry: time.Now().Add(2 * time.Hour),
+		}
+
+		return &apimodel.Content{URL: url}, nil
+	}
+
+	var requestUser string
+
+	if requester != nil {
+		// Set requesting acc username.
+		requestUser = requester.Username
+	}
+
+	// Ensure that stored media is cached.
+	// (this handles local media / recaches).
+	attach, err = p.federator.RefreshMedia(
+		ctx,
+		requestUser,
+		attach,
+		media.AdditionalMediaInfo{},
+		false,
+	)
+	if err != nil {
+		err := gtserror.Newf("error recaching media: %w", err)
+		return nil, gtserror.NewErrorNotFound(err)
+	}
+
+	// Start preparing API content model.
+	apiContent := &apimodel.Content{
+		ContentUpdated: attach.UpdatedAt,
+	}
+
+	// Retrieve appropriate
+	// size file from storage.
+	switch sizeStr {
+
+	case media.SizeOriginal:
+		apiContent.ContentType = attach.File.ContentType
+		apiContent.ContentLength = int64(attach.File.FileSize)
+		return p.getContent(ctx,
+			attach.File.Path,
+			apiContent,
+		)
+
+	case media.SizeSmall:
+		apiContent.ContentType = attach.Thumbnail.ContentType
+		apiContent.ContentLength = int64(attach.Thumbnail.FileSize)
+		return p.getContent(ctx,
+			attach.Thumbnail.Path,
+			apiContent,
+		)
+
+	default:
+		const text = "invalid media attachment size"
+		return nil, gtserror.NewErrorBadRequest(errors.New(text), text)
+	}
+}
+
+func (p *Processor) getEmojiContent(
+	ctx context.Context,
+
+	ownerID string,
+	emojiID string,
+	sizeStr media.Size,
+) (
+	*apimodel.Content,
+	gtserror.WithCode,
+) {
+	// Reconstruct static emoji image URL to search for it.
+	// As refreshed emojis use a newly generated path ID to
+	// differentiate them (cache-wise) from the original.
+	staticURL := uris.URIForAttachment(
+		ownerID,
+		string(media.TypeEmoji),
+		string(media.SizeStatic),
+		emojiID,
+		"png",
+	)
+
+	// Search for emoji with given static URL in the database.
+	emoji, err := p.state.DB.GetEmojiByStaticURL(ctx, staticURL)
+	if err != nil && !errors.Is(err, db.ErrNoEntries) {
+		err := gtserror.Newf("error fetching emoji from database: %w", err)
+		return nil, gtserror.NewErrorInternalError(err)
+	}
+
+	if emoji == nil {
+		const text = "emoji not found"
+		return nil, gtserror.NewErrorNotFound(errors.New(text), text)
+	}
+
+	if *emoji.Disabled {
+		const text = "emoji has been disabled"
+		return nil, gtserror.NewErrorNotFound(errors.New(text), text)
+	}
+
+	// Ensure that stored emoji is cached.
+	// (this handles local emoji / recaches).
+	emoji, err = p.federator.RefreshEmoji(
+		ctx,
+		emoji,
+		media.AdditionalEmojiInfo{},
+		false,
+	)
+	if err != nil {
+		err := gtserror.Newf("error recaching emoji: %w", err)
+		return nil, gtserror.NewErrorNotFound(err)
+	}
+
+	// Start preparing API content model.
+	apiContent := &apimodel.Content{}
+
+	// Retrieve appropriate
+	// size file from storage.
+	switch sizeStr {
+
+	case media.SizeOriginal:
+		apiContent.ContentType = emoji.ImageContentType
+		apiContent.ContentLength = int64(emoji.ImageFileSize)
+		return p.getContent(ctx,
+			emoji.ImagePath,
+			apiContent,
+		)
+
+	case media.SizeStatic:
+		apiContent.ContentType = emoji.ImageStaticContentType
+		apiContent.ContentLength = int64(emoji.ImageStaticFileSize)
+		return p.getContent(ctx,
+			emoji.ImageStaticPath,
+			apiContent,
+		)
+
+	default:
+		const text = "invalid media attachment size"
+		return nil, gtserror.NewErrorBadRequest(errors.New(text), text)
+	}
+}
+
+// getContent performs the final file fetching of
+// stored content at path in storage. This is
+// populated in the apimodel.Content{} and returned.
+// (note: this also handles un-proxied S3 storage).
+func (p *Processor) getContent(
+	ctx context.Context,
+	path string,
+	content *apimodel.Content,
+) (
+	*apimodel.Content,
+	gtserror.WithCode,
+) {
+	// If running on S3 storage with proxying disabled then
+	// just fetch pre-signed URL instead of the content.
+	if url := p.state.Storage.URL(ctx, path); url != nil {
+		content.URL = url
+		return content, nil
+	}
+
+	// Fetch file stream for the stored media at path.
+	rc, err := p.state.Storage.GetStream(ctx, path)
+	if err != nil && !storage.IsNotFound(err) {
+		err := gtserror.Newf("error getting file %s from storage: %w", path, err)
+		return nil, gtserror.NewErrorInternalError(err)
+	}
+
+	// Ensure found.
+	if rc == nil {
+		const text = "file not found"
+		return nil, gtserror.NewErrorNotFound(errors.New(text), text)
+	}
+
+	// Return with stream.
+	content.Content = rc
+	return content, nil
+}
 
 func parseType(s string) (media.Type, error) {
 	switch s {
@@ -120,198 +357,3 @@ func parseSize(s string) (media.Size, error) {
 	}
 	return "", fmt.Errorf("%s not a recognized media.Size", s)
 }
-
-func (p *Processor) getAttachmentContent(ctx context.Context, requestingAccount *gtsmodel.Account, wantedMediaID string, owningAccountID string, mediaSize media.Size) (*apimodel.Content, gtserror.WithCode) {
-	// retrieve attachment from the database and do basic checks on it
-	a, err := p.state.DB.GetAttachmentByID(ctx, wantedMediaID)
-	if err != nil {
-		err = gtserror.Newf("attachment %s could not be taken from the db: %w", wantedMediaID, err)
-		return nil, gtserror.NewErrorNotFound(err)
-	}
-
-	if a.AccountID != owningAccountID {
-		err = gtserror.Newf("attachment %s is not owned by %s", wantedMediaID, owningAccountID)
-		return nil, gtserror.NewErrorNotFound(err)
-	}
-
-	// If this is an "Unknown" file type, ie., one we
-	// tried to process and couldn't, or one we refused
-	// to process because it wasn't supported, then we
-	// can skip a lot of steps here by simply forwarding
-	// the request to the remote URL.
-	if a.Type == gtsmodel.FileTypeUnknown {
-		remoteURL, err := url.Parse(a.RemoteURL)
-		if err != nil {
-			err = gtserror.Newf("error parsing remote URL of 'Unknown'-type attachment for redirection: %w", err)
-			return nil, gtserror.NewErrorInternalError(err)
-		}
-
-		url := &storage.PresignedURL{
-			URL: remoteURL,
-			// We might manage to cache the media
-			// at some point, so set a low-ish expiry.
-			Expiry: time.Now().Add(2 * time.Hour),
-		}
-
-		return &apimodel.Content{URL: url}, nil
-	}
-
-	if !*a.Cached {
-		// if we don't have it cached, then we can assume two things:
-		// 1. this is remote media, since local media should never be uncached
-		// 2. we need to fetch it again using a transport and the media manager
-		remoteMediaIRI, err := url.Parse(a.RemoteURL)
-		if err != nil {
-			return nil, gtserror.NewErrorNotFound(fmt.Errorf("error parsing remote media iri %s: %w", a.RemoteURL, err))
-		}
-
-		// use an empty string as requestingUsername to use the instance account, unless the request for this
-		// media has been http signed, then use the requesting account to make the request to remote server
-		var requestingUsername string
-		if requestingAccount != nil {
-			requestingUsername = requestingAccount.Username
-		}
-
-		// Pour one out for tobi's original streamed recache
-		// (streaming data both to the client and storage).
-		// Gone and forever missed <3
-		//
-		// [
-		//   the reason it was removed was because a slow
-		//   client connection could hold open a storage
-		//   recache operation -> holding open a media worker.
-		// ]
-
-		dataFn := func(ctx context.Context) (io.ReadCloser, int64, error) {
-			t, err := p.transportController.NewTransportForUsername(ctx, requestingUsername)
-			if err != nil {
-				return nil, 0, err
-			}
-			return t.DereferenceMedia(gtscontext.SetFastFail(ctx), remoteMediaIRI)
-		}
-
-		// Start recaching this media with the prepared data function.
-		processingMedia, err := p.mediaManager.PreProcessMediaRecache(ctx, dataFn, wantedMediaID)
-		if err != nil {
-			return nil, gtserror.NewErrorNotFound(fmt.Errorf("error recaching media: %w", err))
-		}
-
-		// Load attachment and block until complete
-		a, err = processingMedia.LoadAttachment(ctx)
-		if err != nil {
-			return nil, gtserror.NewErrorNotFound(fmt.Errorf("error loading recached attachment: %w", err))
-		}
-	}
-
-	var (
-		storagePath       string
-		attachmentContent = &apimodel.Content{
-			ContentUpdated: a.UpdatedAt,
-		}
-	)
-
-	// get file information from the attachment depending on the requested media size
-	switch mediaSize {
-	case media.SizeOriginal:
-		attachmentContent.ContentType = a.File.ContentType
-		attachmentContent.ContentLength = int64(a.File.FileSize)
-		storagePath = a.File.Path
-	case media.SizeSmall:
-		attachmentContent.ContentType = a.Thumbnail.ContentType
-		attachmentContent.ContentLength = int64(a.Thumbnail.FileSize)
-		storagePath = a.Thumbnail.Path
-	default:
-		return nil, gtserror.NewErrorNotFound(fmt.Errorf("media size %s not recognized for attachment", mediaSize))
-	}
-
-	// ... so now we can safely return it
-	return p.retrieveFromStorage(ctx, storagePath, attachmentContent)
-}
-
-func (p *Processor) getEmojiContent(ctx context.Context, fileName string, owningAccountID string, emojiSize media.Size) (*apimodel.Content, gtserror.WithCode) {
-	emojiContent := &apimodel.Content{}
-	var storagePath string
-
-	// reconstruct the static emoji image url -- reason
-	// for using the static URL rather than full size url
-	// is that static emojis are always encoded as png,
-	// so this is more reliable than using full size url
-	imageStaticURL := uris.URIForAttachment(
-		owningAccountID,
-		string(media.TypeEmoji),
-		string(media.SizeStatic),
-		fileName,
-		"png",
-	)
-
-	e, err := p.state.DB.GetEmojiByStaticURL(ctx, imageStaticURL)
-	if err != nil {
-		return nil, gtserror.NewErrorNotFound(fmt.Errorf("emoji %s could not be taken from the db: %w", fileName, err))
-	}
-
-	if *e.Disabled {
-		return nil, gtserror.NewErrorNotFound(fmt.Errorf("emoji %s has been disabled", fileName))
-	}
-
-	if !*e.Cached {
-		// if we don't have it cached, then we can assume two things:
-		// 1. this is remote emoji, since local emoji should never be uncached
-		// 2. we need to fetch it again using a transport and the media manager
-		remoteURL, err := url.Parse(e.ImageRemoteURL)
-		if err != nil {
-			return nil, gtserror.NewErrorNotFound(fmt.Errorf("error parsing remote emoji iri %s: %w", e.ImageRemoteURL, err))
-		}
-
-		dataFn := func(ctx context.Context) (io.ReadCloser, int64, error) {
-			t, err := p.transportController.NewTransportForUsername(ctx, "")
-			if err != nil {
-				return nil, 0, err
-			}
-			return t.DereferenceMedia(gtscontext.SetFastFail(ctx), remoteURL)
-		}
-
-		// Start recaching this emoji with the prepared data function.
-		processingEmoji, err := p.mediaManager.PreProcessEmojiRecache(ctx, dataFn, e.ID)
-		if err != nil {
-			return nil, gtserror.NewErrorNotFound(fmt.Errorf("error recaching emoji: %w", err))
-		}
-
-		// Load attachment and block until complete
-		e, err = processingEmoji.LoadEmoji(ctx)
-		if err != nil {
-			return nil, gtserror.NewErrorNotFound(fmt.Errorf("error loading recached emoji: %w", err))
-		}
-	}
-
-	switch emojiSize {
-	case media.SizeOriginal:
-		emojiContent.ContentType = e.ImageContentType
-		emojiContent.ContentLength = int64(e.ImageFileSize)
-		storagePath = e.ImagePath
-	case media.SizeStatic:
-		emojiContent.ContentType = e.ImageStaticContentType
-		emojiContent.ContentLength = int64(e.ImageStaticFileSize)
-		storagePath = e.ImageStaticPath
-	default:
-		return nil, gtserror.NewErrorNotFound(fmt.Errorf("media size %s not recognized for emoji", emojiSize))
-	}
-
-	return p.retrieveFromStorage(ctx, storagePath, emojiContent)
-}
-
-func (p *Processor) retrieveFromStorage(ctx context.Context, storagePath string, content *apimodel.Content) (*apimodel.Content, gtserror.WithCode) {
-	// If running on S3 storage with proxying disabled then
-	// just fetch a pre-signed URL instead of serving the content.
-	if url := p.state.Storage.URL(ctx, storagePath); url != nil {
-		content.URL = url
-		return content, nil
-	}
-
-	reader, err := p.state.Storage.GetStream(ctx, storagePath)
-	if err != nil {
-		return nil, gtserror.NewErrorNotFound(fmt.Errorf("error retrieving from storage: %s", err))
-	}
-
-	content.Content = reader
-	return content, nil
-}