import "github.com/guregu/dynamo/v2"
dynamo is an expressive DynamoDB client for Go, with an easy but powerful API. dynamo integrates with the official AWS SDK v2.
This library is stable and versioned with Go modules.
Tip
dynamo v2 is finally released! See v2 Migration for tips on migrating from dynamo v1.
For dynamo v1, which uses aws-sdk-go v1, see: dynamo v1 documentation.
package dynamo
import (
"time"
"context"
"log"
"github.com/aws/aws-sdk-go-v2/aws"
"github.com/aws/aws-sdk-go-v2/config"
"github.com/guregu/dynamo/v2"
)
// Use struct tags much like the standard JSON library,
// you can embed anonymous structs too!
type widget struct {
UserID int // Hash key, a.k.a. partition key
Time time.Time // Range key, a.k.a. sort key
Msg string `dynamo:"Message"` // Change name in the database
Count int `dynamo:",omitempty"` // Omits if zero value
Children []widget // List of maps
Friends []string `dynamo:",set"` // Sets
Set map[string]struct{} `dynamo:",set"` // Map sets, too!
SecretKey string `dynamo:"-"` // Ignored
}
func main() {
cfg, err := config.LoadDefaultConfig(context.TODO(), config.WithRegion("us-east-1"))
if err != nil {
log.Fatalf("unable to load SDK config, %v", err)
}
db := dynamo.New(cfg)
table := db.Table("Widgets")
// put item
w := widget{UserID: 613, Time: time.Now(), Msg: "hello"}
err = table.Put(w).Run(ctx)
// get the same item
var result widget
err = table.Get("UserID", w.UserID).
Range("Time", dynamo.Equal, w.Time).
One(ctx, &result)
// get all items
var results []widget
err = table.Scan().All(ctx, &results)
// use placeholders in filter expressions (see Expressions section below)
var filtered []widget
err = table.Scan().Filter("'Count' > ?", 10).All(ctx, &filtered)
}
dynamo will help you write expressions used to filter results in queries and scans, and add conditions to puts and deletes.
Attribute names may be written as is if it is not a reserved word, or be escaped with single quotes (''
). You may also use dollar signs ($
) as placeholders for attribute names and list indexes. DynamoDB has very large amount of reserved words so it may be a good idea to just escape everything.
Question marks (?
) are used as placeholders for attribute values. DynamoDB doesn't have value literals, so you need to substitute everything.
Please see the DynamoDB reference on expressions for more information. The Comparison Operator and Function Reference is also handy.
// Using single quotes to escape a reserved word, and a question mark as a value placeholder.
// Finds all items whose date is greater than or equal to lastUpdate.
table.Scan().Filter("'Date' >= ?", lastUpdate).All(ctx, &results)
// Using dollar signs as a placeholder for attribute names.
// Deletes the item with an ID of 42 if its score is at or below the cutoff, and its name starts with G.
table.Delete("ID", 42).If("Score <= ? AND begins_with($, ?)", cutoff, "Name", "G").Run(ctx)
// Put a new item, only if it doesn't already exist.
table.Put(item{ID: 42}).If("attribute_not_exists(ID)").Run(ctx)
dynamo automatically handles the following interfaces:
dynamo.Marshaler
anddynamo.Unmarshaler
dynamodbattribute.Marshaler
anddynamodbattribute.Unmarshaler
encoding.TextMarshaler
andencoding.TextUnmarshaler
This allows you to define custom encodings and provides built-in support for types such as time.Time
.
dynamo handles struct tags similarly to the standard library encoding/json
package. It uses dynamo
for the struct tag's name, taking the form of: dynamo:"attributeName,option1,option2,etc"
. You can omit the attribute name to use the default: dynamo:",option1,etc"
.
By default, dynamo will use the name of your fields as the name of the DynamoDB attribute it corresponds do. You can specify a different name with the dynamo
struct tag like so: dynamo:"other_name_goes_here"
. If two fields have the same name, dynamo will prioritize the higher-level field.
If you set a field's name to "-"
(as in dynamo:"-"
) that field will be ignored. It will be omitted when marshaling and ignored when unmarshaling. Also, fields that start with a lowercase letter will be ignored. However, embedding a struct whose type has a lowercase letter but contains uppercase fields is OK.
By default, slices will be marshaled as DynamoDB lists. To marshal a field to sets instead, use the dynamo:",set"
option. Empty sets will be automatically omitted.
You can use maps as sets too. The following types are supported:
[]T
map[T]struct{}
map[T]bool
where T
represents any type that marshals into a DynamoDB string, number, or binary value.
Note that the order of objects within a set is undefined.
Using the omitempty option (as in dynamo:",omitempty"
) will omit the field if it has a zero (ex. an empty string, 0, nil pointer) value. Structs are supported.
It also supports the isZeroer
interface below:
type isZeroer interface {
IsZero() bool
}
If IsZero()
returns true, the field will be omitted. This gives us built-in support for time.Time
.
You can also use the dynamo:",omitemptyelem"
option to omit empty values inside of slices.
Some values will be automatically omitted.
- Empty strings
- Empty sets
- Empty structs
- Nil pointers and interfaces
- Types that implement
encoding.TextMarshaler
and whoseMarshalText
method returns 0-length or nil slice. - Zero-length binary (byte slices)
To override this behavior, use the dynamo:",allowempty"
flag. Not all empty types can be stored by DynamoDB. For example, empty sets will still be omitted.
To override auto-omit behavior for children of a map, for example map[string]string
, use the dynamo:",allowemptyelem"
option.
DynamoDB has a special NULL type to represent null values. In general, this library avoids marshaling things as NULL and prefers to omit those values instead. If you want empty/nil values to marshal to NULL, use the dynamo:",null"
option.
By default, time.Time
will marshal to a string because it implements encoding.TextMarshaler
.
If you want time.Time
to marshal as a Unix time value (number of seconds since the Unix epoch), you can use the dynamo:",unixtime"
option. This is useful for TTL fields, which must be Unix time.
You can use struct tags to specify hash keys, range keys, and indexes when creating a table.
For example:
type UserAction struct {
UserID string `dynamo:"ID,hash" index:"Seq-ID-index,range"`
Time time.Time `dynamo:",range"`
Seq int64 `localIndex:"ID-Seq-index,range" index:"Seq-ID-index,hash"`
UUID string `index:"UUID-index,hash"`
}
This creates a table with the primary hash key ID and range key Time. It creates two global secondary indices called UUID-index and Seq-ID-index, and a local secondary index called ID-Seq-index.
As of v2, dynamo relies on the AWS SDK for retrying. See: Retries and Timeouts documentation for information about how to configure its behavior.
By default, canceled transactions (i.e. errors from conflicting transactions) will not be retried. To get automatic retrying behavior like in v1, use dynamo.RetryTxConflicts
.
import (
"context"
"log"
"github.com/aws/aws-sdk-go-v2/aws"
"github.com/aws/aws-sdk-go-v2/aws/retry"
"github.com/aws/aws-sdk-go-v2/config"
"github.com/guregu/dynamo/v2"
)
func main() {
cfg, err := config.LoadDefaultConfig(context.Background(), config.WithRetryer(func() aws.Retryer {
return retry.NewStandard(dynamo.RetryTxConflicts)
}))
if err != nil {
log.Fatal(err)
}
db := dynamo.New(cfg)
// use db
}
dynamo has been in development before the official AWS libraries were stable. We use a different encoder and decoder than the dynamodbattribute package. dynamo uses the dynamo
struct tag instead of the dynamodbav
struct tag, and we also prefer to automatically omit invalid values such as empty strings, whereas the dynamodbattribute package substitutes null values for them. Items that satisfy the dynamodbattribute.(Un)marshaler
interfaces are compatibile with both libraries.
In order to use dynamodbattribute's encoding facilities, you must wrap objects passed to dynamo with dynamo.AWSEncoding
. Here is a quick example:
// Notice the use of the dynamodbav struct tag
type book struct {
ID int `dynamodbav:"id"`
Title string `dynamodbav:"title"`
}
// Putting an item
err := db.Table("Books").Put(dynamo.AWSEncoding(book{
ID: 42,
Title: "Principia Discordia",
})).Run(ctx)
// When getting an item you MUST pass a pointer to AWSEncoding!
var someBook book
err := db.Table("Books").Get("ID", 555).One(ctx, dynamo.AWSEncoding(&someBook))
The API hasn't changed much from v1 to v2. Here are some migration tips:
- All request methods now take a context as their first argument.
- Retrying relies on the AWS SDK configuration, see: Retrying.
- Transactions won't retry TransactionCanceled responses by default anymore, make sure you configure that if you need it.
- Arguments that took
int64
(such as inQuery.Limit
) now takeint
instead. - Compatibility with the official AWS library uses v2 interfaces instead of v1.
KMSMasterKeyArn
renamed toKMSMasterKeyARN
.
By default, tests are run in offline mode. In order to run the integration tests, some environment variables need to be set.
To run the tests against DynamoDB Local:
# Use Docker to run DynamoDB local on port 8880
docker compose -f '.github/docker-compose.yml' up -d
# Run the tests with a fresh table
# The tables will be created automatically
# The '%' in the table name will be replaced the current timestamp
DYNAMO_TEST_ENDPOINT='http://localhost:8880' \
DYNAMO_TEST_REGION='local' \
DYNAMO_TEST_TABLE='TestDB-%' \
AWS_ACCESS_KEY_ID='dummy' \
AWS_SECRET_ACCESS_KEY='dummy' \
AWS_REGION='local' \
go test -v -race ./... -cover -coverpkg=./...
BSD 2-Clause