[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
This commit is contained in:
kim 2024-06-26 15:01:16 +00:00 committed by GitHub
commit 21bb324156
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
48 changed files with 2578 additions and 1926 deletions

View file

@ -19,29 +19,190 @@ package dereferencing
import (
"context"
"fmt"
"errors"
"io"
"net/url"
"github.com/superseriousbusiness/gotosocial/internal/db"
"github.com/superseriousbusiness/gotosocial/internal/gtserror"
"github.com/superseriousbusiness/gotosocial/internal/gtsmodel"
"github.com/superseriousbusiness/gotosocial/internal/id"
"github.com/superseriousbusiness/gotosocial/internal/log"
"github.com/superseriousbusiness/gotosocial/internal/media"
"github.com/superseriousbusiness/gotosocial/internal/util"
)
func (d *Dereferencer) GetRemoteEmoji(ctx context.Context, requestUser string, remoteURL string, shortcode string, domain string, id string, emojiURI string, ai *media.AdditionalEmojiInfo, refresh bool) (*media.ProcessingEmoji, error) {
var shortcodeDomain = shortcode + "@" + domain
// Ensure we have been passed a valid URL.
derefURI, err := url.Parse(remoteURL)
if err != nil {
return nil, fmt.Errorf("GetRemoteEmoji: error parsing url for emoji %s: %s", shortcodeDomain, err)
// GetEmoji fetches the emoji with given shortcode,
// domain and remote URL to dereference it by. This
// handles the case of existing emojis by passing them
// to RefreshEmoji(), which in the case of a local
// emoji will be a no-op. If the emoji does not yet
// exist it will be newly inserted into the database
// followed by dereferencing the actual media file.
//
// Please note that even if an error is returned,
// an emoji model may still be returned if the error
// was only encountered during actual dereferencing.
// In this case, it will act as a placeholder.
func (d *Dereferencer) GetEmoji(
ctx context.Context,
shortcode string,
domain string,
remoteURL string,
info media.AdditionalEmojiInfo,
refresh bool,
) (
*gtsmodel.Emoji,
error,
) {
// Look for an existing emoji with shortcode domain.
emoji, err := d.state.DB.GetEmojiByShortcodeDomain(ctx,
shortcode,
domain,
)
if err != nil && !errors.Is(err, db.ErrNoEntries) {
return nil, gtserror.Newf("error fetching emoji from db: %w", err)
}
// Acquire derefs lock.
if emoji != nil {
// This was an existing emoji, pass to refresh func.
return d.RefreshEmoji(ctx, emoji, info, refresh)
}
if domain == "" {
// failed local lookup, will be db.ErrNoEntries.
return nil, gtserror.SetUnretrievable(err)
}
// Generate shortcode domain for locks + logging.
shortcodeDomain := shortcode + "@" + domain
// Ensure we have a valid remote URL.
url, err := url.Parse(remoteURL)
if err != nil {
err := gtserror.Newf("invalid image remote url %s for emoji %s: %w", remoteURL, shortcodeDomain, err)
return nil, err
}
// Acquire new instance account transport for emoji dereferencing.
tsport, err := d.transportController.NewTransportForUsername(ctx, "")
if err != nil {
err := gtserror.Newf("error getting instance transport: %w", err)
return nil, err
}
// Prepare data function to dereference remote emoji media.
data := func(context.Context) (io.ReadCloser, int64, error) {
return tsport.DereferenceMedia(ctx, url)
}
// Pass along for safe processing.
return d.processEmojiSafely(ctx,
shortcodeDomain,
func() (*media.ProcessingEmoji, error) {
return d.mediaManager.CreateEmoji(ctx,
shortcode,
domain,
data,
info,
)
},
)
}
// RefreshEmoji ensures that the given emoji is
// up-to-date, both in terms of being cached in
// in local instance storage, and compared to extra
// information provided in media.AdditionEmojiInfo{}.
// (note that is a no-op to pass in a local emoji).
//
// Please note that even if an error is returned,
// an emoji model may still be returned if the error
// was only encountered during actual dereferencing.
// In this case, it will act as a placeholder.
func (d *Dereferencer) RefreshEmoji(
ctx context.Context,
emoji *gtsmodel.Emoji,
info media.AdditionalEmojiInfo,
force bool,
) (
*gtsmodel.Emoji,
error,
) {
// Can't refresh local.
if emoji.IsLocal() {
return emoji, nil
}
// Check emoji is up-to-date
// with provided extra info.
switch {
case info.URI != nil &&
*info.URI != emoji.URI:
force = true
case info.ImageRemoteURL != nil &&
*info.ImageRemoteURL != emoji.ImageRemoteURL:
force = true
case info.ImageStaticRemoteURL != nil &&
*info.ImageStaticRemoteURL != emoji.ImageStaticRemoteURL:
force = true
}
// Check if needs updating.
if !force && *emoji.Cached {
return emoji, nil
}
// TODO: more finegrained freshness checks.
// Generate shortcode domain for locks + logging.
shortcodeDomain := emoji.Shortcode + "@" + emoji.Domain
// Ensure we have a valid image remote URL.
url, err := url.Parse(emoji.ImageRemoteURL)
if err != nil {
err := gtserror.Newf("invalid image remote url %s for emoji %s: %w", emoji.ImageRemoteURL, shortcodeDomain, err)
return nil, err
}
// Acquire new instance account transport for emoji dereferencing.
tsport, err := d.transportController.NewTransportForUsername(ctx, "")
if err != nil {
err := gtserror.Newf("error getting instance transport: %w", err)
return nil, err
}
// Prepare data function to dereference remote emoji media.
data := func(context.Context) (io.ReadCloser, int64, error) {
return tsport.DereferenceMedia(ctx, url)
}
// Pass along for safe processing.
return d.processEmojiSafely(ctx,
shortcodeDomain,
func() (*media.ProcessingEmoji, error) {
return d.mediaManager.RefreshEmoji(ctx,
emoji,
data,
info,
)
},
)
}
// processingEmojiSafely provides concurrency-safe processing of
// an emoji with given shortcode+domain. if a copy of the emoji is
// not already being processed, the given 'process' callback will
// be used to generate new *media.ProcessingEmoji{} instance.
func (d *Dereferencer) processEmojiSafely(
ctx context.Context,
shortcodeDomain string,
process func() (*media.ProcessingEmoji, error),
) (
emoji *gtsmodel.Emoji,
err error,
) {
// Acquire map lock.
d.derefEmojisMu.Lock()
// Ensure unlock only done once.
@ -53,146 +214,118 @@ func (d *Dereferencer) GetRemoteEmoji(ctx context.Context, requestUser string, r
processing, ok := d.derefEmojis[shortcodeDomain]
if !ok {
// Fetch a transport for current request user in order to perform request.
tsport, err := d.transportController.NewTransportForUsername(ctx, requestUser)
// Start new processing emoji.
processing, err = process()
if err != nil {
return nil, gtserror.Newf("couldn't create transport: %w", err)
return nil, err
}
// Set the media data function to dereference emoji from URI.
data := func(ctx context.Context) (io.ReadCloser, int64, error) {
return tsport.DereferenceMedia(ctx, derefURI)
}
// Create new emoji processing request from the media manager.
processing, err = d.mediaManager.PreProcessEmoji(ctx, data,
shortcode,
id,
emojiURI,
ai,
refresh,
)
if err != nil {
return nil, gtserror.Newf("error preprocessing emoji %s: %s", shortcodeDomain, err)
}
// Store media in map to mark as processing.
d.derefEmojis[shortcodeDomain] = processing
defer func() {
// On exit safely remove emoji from map.
d.derefEmojisMu.Lock()
delete(d.derefEmojis, shortcodeDomain)
d.derefEmojisMu.Unlock()
}()
}
// Unlock map.
unlock()
// Start emoji attachment loading (blocking call).
if _, err := processing.LoadEmoji(ctx); err != nil {
return nil, err
// Perform emoji load operation.
emoji, err = processing.Load(ctx)
if err != nil {
err = gtserror.Newf("error loading emoji %s: %w", shortcodeDomain, err)
// TODO: in time we should return checkable flags by gtserror.Is___()
// which can determine if loading error should allow remaining placeholder.
}
return processing, nil
// Return a COPY of emoji.
emoji2 := new(gtsmodel.Emoji)
*emoji2 = *emoji
return emoji2, err
}
func (d *Dereferencer) populateEmojis(ctx context.Context, rawEmojis []*gtsmodel.Emoji, requestingUsername string) ([]*gtsmodel.Emoji, error) {
// At this point we should know:
// * the AP uri of the emoji
// * the domain of the emoji
// * the shortcode of the emoji
// * the remote URL of the image
// This should be enough to dereference the emoji
gotEmojis := make([]*gtsmodel.Emoji, 0, len(rawEmojis))
func (d *Dereferencer) fetchEmojis(
ctx context.Context,
existing []*gtsmodel.Emoji,
emojis []*gtsmodel.Emoji, // newly dereferenced
) (
[]*gtsmodel.Emoji,
bool, // any changes?
error,
) {
// Track any changes.
changed := false
for _, e := range rawEmojis {
var gotEmoji *gtsmodel.Emoji
var err error
shortcodeDomain := e.Shortcode + "@" + e.Domain
for i, placeholder := range emojis {
// Look for an existing emoji with shortcode + domain.
existing, ok := getEmojiByShortcodeDomain(existing,
placeholder.Shortcode,
placeholder.Domain,
)
if ok && existing.ID != "" {
// check if we already know this emoji
if e.ID != "" {
// we had an ID for this emoji already, which means
// it should be fleshed out already and we won't
// have to get it from the database again
gotEmoji = e
} else if gotEmoji, err = d.state.DB.GetEmojiByShortcodeDomain(ctx, e.Shortcode, e.Domain); err != nil && err != db.ErrNoEntries {
log.Errorf(ctx, "error checking database for emoji %s: %s", shortcodeDomain, err)
// Check for any emoji changes that
// indicate we should force a refresh.
force := emojiChanged(existing, placeholder)
// Ensure that the existing emoji model is up-to-date and cached.
existing, err := d.RefreshEmoji(ctx, existing, media.AdditionalEmojiInfo{
// Set latest values from placeholder.
URI: &placeholder.URI,
ImageRemoteURL: &placeholder.ImageRemoteURL,
ImageStaticRemoteURL: &placeholder.ImageStaticRemoteURL,
}, force)
if err != nil {
log.Errorf(ctx, "error refreshing emoji: %v", err)
// specifically do NOT continue here,
// we already have a model, we don't
// want to drop it from the slice, just
// log that an update for it failed.
}
// Set existing emoji.
emojis[i] = existing
continue
}
var refresh bool
// Emojis changed!
changed = true
if gotEmoji != nil {
// we had the emoji already, but refresh it if necessary
if e.UpdatedAt.Unix() > gotEmoji.ImageUpdatedAt.Unix() {
log.Tracef(ctx, "emoji %s was updated since we last saw it, will refresh", shortcodeDomain)
refresh = true
}
if !refresh && (e.URI != gotEmoji.URI) {
log.Tracef(ctx, "emoji %s changed URI since we last saw it, will refresh", shortcodeDomain)
refresh = true
}
if !refresh && (e.ImageRemoteURL != gotEmoji.ImageRemoteURL) {
log.Tracef(ctx, "emoji %s changed image URL since we last saw it, will refresh", shortcodeDomain)
refresh = true
}
if !refresh {
log.Tracef(ctx, "emoji %s is up to date, will not refresh", shortcodeDomain)
} else {
log.Tracef(ctx, "refreshing emoji %s", shortcodeDomain)
emojiID := gotEmoji.ID // use existing ID
processingEmoji, err := d.GetRemoteEmoji(ctx, requestingUsername, e.ImageRemoteURL, e.Shortcode, e.Domain, emojiID, e.URI, &media.AdditionalEmojiInfo{
Domain: &e.Domain,
ImageRemoteURL: &e.ImageRemoteURL,
ImageStaticRemoteURL: &e.ImageStaticRemoteURL,
Disabled: gotEmoji.Disabled,
VisibleInPicker: gotEmoji.VisibleInPicker,
}, refresh)
if err != nil {
log.Errorf(ctx, "couldn't refresh remote emoji %s: %s", shortcodeDomain, err)
continue
}
if gotEmoji, err = processingEmoji.LoadEmoji(ctx); err != nil {
log.Errorf(ctx, "couldn't load refreshed remote emoji %s: %s", shortcodeDomain, err)
continue
}
}
} else {
// it's new! go get it!
newEmojiID, err := id.NewRandomULID()
if err != nil {
log.Errorf(ctx, "error generating id for remote emoji %s: %s", shortcodeDomain, err)
// Fetch this newly added emoji,
// this function handles the case
// of existing cached emojis and
// new ones requiring dereference.
emoji, err := d.GetEmoji(ctx,
placeholder.Shortcode,
placeholder.Domain,
placeholder.ImageRemoteURL,
media.AdditionalEmojiInfo{
URI: &placeholder.URI,
ImageRemoteURL: &placeholder.ImageRemoteURL,
ImageStaticRemoteURL: &placeholder.ImageStaticRemoteURL,
},
false,
)
if err != nil {
if emoji == nil {
log.Errorf(ctx, "error loading emoji %s: %v", placeholder.ImageRemoteURL, err)
continue
}
processingEmoji, err := d.GetRemoteEmoji(ctx, requestingUsername, e.ImageRemoteURL, e.Shortcode, e.Domain, newEmojiID, e.URI, &media.AdditionalEmojiInfo{
Domain: &e.Domain,
ImageRemoteURL: &e.ImageRemoteURL,
ImageStaticRemoteURL: &e.ImageStaticRemoteURL,
Disabled: e.Disabled,
VisibleInPicker: e.VisibleInPicker,
}, refresh)
if err != nil {
log.Errorf(ctx, "couldn't get remote emoji %s: %s", shortcodeDomain, err)
continue
}
if gotEmoji, err = processingEmoji.LoadEmoji(ctx); err != nil {
log.Errorf(ctx, "couldn't load remote emoji %s: %s", shortcodeDomain, err)
continue
}
// non-fatal error occurred during loading, still use it.
log.Warnf(ctx, "partially loaded emoji: %v", err)
}
// if we get here, we either had the emoji already or we successfully fetched it
gotEmojis = append(gotEmojis, gotEmoji)
// Set updated emoji.
emojis[i] = emoji
}
return gotEmojis, nil
for i := 0; i < len(emojis); {
if emojis[i].ID == "" {
// Remove failed emoji populations.
copy(emojis[i:], emojis[i+1:])
emojis = emojis[:len(emojis)-1]
continue
}
i++
}
return emojis, changed, nil
}