MM-13664 Cache external link metadata when populating post metadata (#10128)

* MM-13664 Added LinkMetadata types

* MM-13664 Use LinkMetadata when populating post metadata

* Fix unused import

* Fix index name on SQLite

* Finish adding unit tests

* Address feedback

* Increase max length of URL column to 2048 characters
Этот коммит содержится в:
Harrison Healey
2019-01-28 10:25:08 -05:00
коммит произвёл Carlos Tadeu Panato Junior
родитель 5c76e90a83
Коммит 26684716aa
22 изменённых файлов: 1483 добавлений и 50 удалений

148
model/link_metadata.go Обычный файл
Просмотреть файл

@@ -0,0 +1,148 @@
// Copyright (c) 2015-present Mattermost, Inc. All Rights Reserved.
// See License.txt for license information.
package model
import (
"encoding/binary"
"encoding/json"
"hash/fnv"
"net/http"
"time"
"github.com/dyatlov/go-opengraph/opengraph"
)
const (
LINK_METADATA_TYPE_IMAGE LinkMetadataType = "image"
LINK_METADATA_TYPE_NONE LinkMetadataType = "none"
LINK_METADATA_TYPE_OPENGRAPH LinkMetadataType = "opengraph"
)
type LinkMetadataType string
// LinkMetadata stores arbitrary data about a link posted in a message. This includes dimensions of linked images
// and OpenGraph metadata.
type LinkMetadata struct {
// Hash is a value computed from the URL and Timestamp for use as a primary key in the database.
Hash int64
URL string
Timestamp int64
Type LinkMetadataType
// Data is the actual metadata for the link. It should contain data of one of the following types:
// - *model.PostImage if the linked content is an image
// - *opengraph.OpenGraph if the linked content is an HTML document
// - nil if the linked content has no metadata
Data interface{}
}
func (o *LinkMetadata) PreSave() {
o.Hash = GenerateLinkMetadataHash(o.URL, o.Timestamp)
}
func (o *LinkMetadata) IsValid() *AppError {
if o.URL == "" {
return NewAppError("LinkMetadata.IsValid", "model.link_metadata.is_valid.url.app_error", nil, "", http.StatusBadRequest)
}
if o.Timestamp == 0 || !isRoundedToNearestHour(o.Timestamp) {
return NewAppError("LinkMetadata.IsValid", "model.link_metadata.is_valid.timestamp.app_error", nil, "", http.StatusBadRequest)
}
switch o.Type {
case LINK_METADATA_TYPE_IMAGE:
if o.Data == nil {
return NewAppError("LinkMetadata.IsValid", "model.link_metadata.is_valid.data.app_error", nil, "", http.StatusBadRequest)
}
if _, ok := o.Data.(*PostImage); !ok {
return NewAppError("LinkMetadata.IsValid", "model.link_metadata.is_valid.data_type.app_error", nil, "", http.StatusBadRequest)
}
case LINK_METADATA_TYPE_NONE:
if o.Data != nil {
return NewAppError("LinkMetadata.IsValid", "model.link_metadata.is_valid.data_type.app_error", nil, "", http.StatusBadRequest)
}
case LINK_METADATA_TYPE_OPENGRAPH:
if o.Data == nil {
return NewAppError("LinkMetadata.IsValid", "model.link_metadata.is_valid.data.app_error", nil, "", http.StatusBadRequest)
}
if _, ok := o.Data.(*opengraph.OpenGraph); !ok {
return NewAppError("LinkMetadata.IsValid", "model.link_metadata.is_valid.data_type.app_error", nil, "", http.StatusBadRequest)
}
default:
return NewAppError("LinkMetadata.IsValid", "model.link_metadata.is_valid.type.app_error", nil, "", http.StatusBadRequest)
}
return nil
}
// DeserializeDataToConcreteType converts o.Data from JSON into properly structured data. This is intended to be used
// after getting a LinkMetadata object that has been stored in the database.
func (o *LinkMetadata) DeserializeDataToConcreteType() error {
var b []byte
switch t := o.Data.(type) {
case []byte:
// MySQL uses a byte slice for JSON
b = t
case string:
// Postgres uses a string for JSON
b = []byte(t)
}
if b == nil {
// Data doesn't need to be fixed
return nil
}
var data interface{}
var err error
switch o.Type {
case LINK_METADATA_TYPE_IMAGE:
image := &PostImage{}
err = json.Unmarshal(b, &image)
data = image
case LINK_METADATA_TYPE_OPENGRAPH:
og := &opengraph.OpenGraph{}
json.Unmarshal(b, &og)
data = og
}
if err != nil {
return err
}
o.Data = data
return nil
}
// FloorToNearestHour takes a timestamp (in milliseconds) and returns it rounded to the previous hour in UTC.
func FloorToNearestHour(ms int64) int64 {
t := time.Unix(0, ms*int64(1000*1000))
return time.Date(t.Year(), t.Month(), t.Day(), t.Hour(), 0, 0, 0, t.Location()).UnixNano() / int64(time.Millisecond)
}
// isRoundedToNearestHour returns true if the given timestamp (in milliseconds) has been rounded to the nearest hour in UTC.
func isRoundedToNearestHour(ms int64) bool {
return FloorToNearestHour(ms) == ms
}
// GenerateLinkMetadataHash generates a unique hash for a given URL and timestamp for use as a database key.
func GenerateLinkMetadataHash(url string, timestamp int64) int64 {
hash := fnv.New32()
// Note that we ignore write errors here because the Hash interface says that its Write will never return an error
binary.Write(hash, binary.LittleEndian, timestamp)
hash.Write([]byte(url))
return int64(hash.Sum32())
}

227
model/link_metadata_test.go Обычный файл
Просмотреть файл

@@ -0,0 +1,227 @@
// Copyright (c) 2015-present Mattermost, Inc. All Rights Reserved.
// See License.txt for license information.
package model
import (
"encoding/json"
"testing"
"github.com/dyatlov/go-opengraph/opengraph"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
)
func TestLinkMetadataIsValid(t *testing.T) {
for _, test := range []struct {
Name string
Metadata *LinkMetadata
Expected bool
}{
{
Name: "should be valid image metadata",
Metadata: &LinkMetadata{
URL: "http://example.com",
Timestamp: 1546300800000,
Type: LINK_METADATA_TYPE_IMAGE,
Data: &PostImage{},
},
Expected: true,
},
{
Name: "should be valid opengraph metadata",
Metadata: &LinkMetadata{
URL: "http://example.com",
Timestamp: 1546300800000,
Type: LINK_METADATA_TYPE_OPENGRAPH,
Data: &opengraph.OpenGraph{},
},
Expected: true,
},
{
Name: "should be valid with no metadata",
Metadata: &LinkMetadata{
URL: "http://example.com",
Timestamp: 1546300800000,
Type: LINK_METADATA_TYPE_NONE,
Data: nil,
},
Expected: true,
},
{
Name: "should be invalid because of empty URL",
Metadata: &LinkMetadata{
Timestamp: 1546300800000,
Type: LINK_METADATA_TYPE_IMAGE,
Data: &PostImage{},
},
Expected: false,
},
{
Name: "should be invalid because of empty timestamp",
Metadata: &LinkMetadata{
URL: "http://example.com",
Type: LINK_METADATA_TYPE_IMAGE,
Data: &PostImage{},
},
Expected: false,
},
{
Name: "should be invalid because of unrounded timestamp",
Metadata: &LinkMetadata{
URL: "http://example.com",
Timestamp: 1546300800001,
Type: LINK_METADATA_TYPE_IMAGE,
Data: &PostImage{},
},
Expected: false,
},
{
Name: "should be invalid because of invalid type",
Metadata: &LinkMetadata{
URL: "http://example.com",
Timestamp: 1546300800000,
Type: "garbage",
Data: &PostImage{},
},
Expected: false,
},
{
Name: "should be invalid because of empty data",
Metadata: &LinkMetadata{
URL: "http://example.com",
Timestamp: 1546300800000,
Type: LINK_METADATA_TYPE_IMAGE,
},
Expected: false,
},
{
Name: "should be invalid because of mismatched data and type, image type and opengraph data",
Metadata: &LinkMetadata{
URL: "http://example.com",
Timestamp: 1546300800000,
Type: LINK_METADATA_TYPE_IMAGE,
Data: &opengraph.OpenGraph{},
},
Expected: false,
},
{
Name: "should be invalid because of mismatched data and type, opengraph type and image data",
Metadata: &LinkMetadata{
URL: "http://example.com",
Timestamp: 1546300800000,
Type: LINK_METADATA_TYPE_OPENGRAPH,
Data: &PostImage{},
},
Expected: false,
},
{
Name: "should be invalid because of mismatched data and type, image type and random data",
Metadata: &LinkMetadata{
URL: "http://example.com",
Timestamp: 1546300800000,
Type: LINK_METADATA_TYPE_OPENGRAPH,
Data: &Channel{},
},
Expected: false,
},
} {
t.Run(test.Name, func(t *testing.T) {
err := test.Metadata.IsValid()
if test.Expected {
assert.Nil(t, err)
} else {
assert.NotNil(t, err)
}
})
}
}
func TestLinkMetadataDeserializeDataToConcreteType(t *testing.T) {
t.Run("should convert []byte to PostImage", func(t *testing.T) {
image := &PostImage{
Height: 400,
Width: 500,
}
metadata := &LinkMetadata{
Type: LINK_METADATA_TYPE_IMAGE,
Data: []byte(image.ToJson()),
}
require.IsType(t, []byte{}, metadata.Data)
err := metadata.DeserializeDataToConcreteType()
assert.Nil(t, err)
assert.IsType(t, &PostImage{}, metadata.Data)
assert.Equal(t, *image, *metadata.Data.(*PostImage))
})
t.Run("should convert string to OpenGraph", func(t *testing.T) {
og := &opengraph.OpenGraph{
URL: "http://example.com",
Description: "Hello, world!",
Images: []*opengraph.Image{
{
URL: "http://example.com/image.png",
},
},
}
b, err := json.Marshal(og)
require.Nil(t, err)
metadata := &LinkMetadata{
Type: LINK_METADATA_TYPE_OPENGRAPH,
Data: b,
}
require.IsType(t, []byte{}, metadata.Data)
err = metadata.DeserializeDataToConcreteType()
assert.Nil(t, err)
assert.IsType(t, &opengraph.OpenGraph{}, metadata.Data)
assert.Equal(t, *og, *metadata.Data.(*opengraph.OpenGraph))
})
t.Run("should ignore data of the correct type", func(t *testing.T) {
metadata := &LinkMetadata{
Type: LINK_METADATA_TYPE_OPENGRAPH,
Data: 1234,
}
err := metadata.DeserializeDataToConcreteType()
assert.Nil(t, err)
})
t.Run("should ignore an invalid type", func(t *testing.T) {
metadata := &LinkMetadata{
Type: "garbage",
Data: "garbage",
}
err := metadata.DeserializeDataToConcreteType()
assert.Nil(t, err)
})
t.Run("should return error for invalid data", func(t *testing.T) {
metadata := &LinkMetadata{
Type: LINK_METADATA_TYPE_IMAGE,
Data: "garbage",
}
err := metadata.DeserializeDataToConcreteType()
assert.NotNil(t, err)
})
}
func TestFloorToNearestHour(t *testing.T) {
assert.True(t, isRoundedToNearestHour(FloorToNearestHour(1546346096000)))
}

Просмотреть файл

@@ -3,6 +3,10 @@
package model
import (
"encoding/json"
)
type PostMetadata struct {
// Embeds holds information required to render content embedded in the post. This includes the OpenGraph metadata
// for links in the post.
@@ -28,3 +32,8 @@ type PostImage struct {
Width int `json:"width"`
Height int `json:"height"`
}
func (o *PostImage) ToJson() string {
b, _ := json.Marshal(o)
return string(b)
}