An Atom is a simple generic data container for holding any type of Plain
Old Data (POD). An Atom can contain simple primitive types like integers,
floating point numbers, and strings; as well as structured data like lists and
dictionary-like Objects
. Since Atoms are POD, they can be easily copied
(for example, with memcpy()
) anywhere and are suitable for use in real-time
code.
Every atom starts with an LV2_Atom header, followed by the contents. This allows code to process atoms without requiring special code for every type of data. For example, plugins that mutually understand a type can be used together in a host that does not understand that type, because the host is only required to copy atoms, not interpret their contents. Similarly, plugins (such as routers, delays, or data structures) can meaningfully process atoms of a type unknown to them.
Atoms should be used anywhere values of various types must be stored or transmitted. An AtomPort can be used to transmit atoms via ports. An AtomPort that contains a Sequence can be used for sample accurate communication of events, such as MIDI.
Serialisation
Each Atom type defines a binary format for use at runtime, but also a serialisation that is natural to express in Turtle format. Thus, this specification defines a powerful real-time appropriate data model, as well as a portable way to serialise any data in that model. This is particularly useful for inter-process communication, saving/restoring state, and describing values in plugin data files.
Custom Atom Types
While it is possible to define new Atom types for any binary format, the
standard types defined here are powerful enough to describe almost anything.
Implementations SHOULD build structures out of the types provided here, rather
than define new binary formats (for example, using Object rather than a
new C struct
type). Host and tool implementations have support for
serialising all standard types, so new binary formats are an implementation
burden which harms interoperabilty. In particular, plugins SHOULD NOT expect
UI communication or state saving with custom binary types to work. In general,
new Atom types should only be defined where absolutely necessary due to
performance reasons and serialisation is not a concern.
Index
Classes
Atom
ClassAtomPort
ClassLabel | Atom Port |
---|---|
Subclass of | lv2:Port |
In domain of | bufferType |
A port which contains an Atom.
Ports of this type are connected to an LV2_Atom with a type specified by bufferType.
Output ports with a variably sized type MUST be initialised by the host before every run() to an Chunk with size set to the available space. The plugin reads this size to know how much space is available for writing. In all cases, the plugin MUST write a complete atom (including header) to outputs. However, to be robust, hosts SHOULD initialise output ports to a safe sentinel (e.g. the null Atom) before calling run().
Blank
ClassLabel | Blank |
---|---|
Subclass of | Object |
cType | LV2_Atom_Object |
owl:deprecated | true |
An anonymous collection of properties without a URI.
This class is deprecated. Use Object with ID 0 instead.
An Object where the LV2_Atom_Object::id is a blank node ID (NOT a URI). The ID of a Blank is valid only within the context the Blank appears in. For ports this is the context of the associated run() call, i.e. all ports share the same context so outputs can contain IDs that correspond to IDs of blanks in the input.
Bool
ClassLabel | Bool |
---|---|
Subclass of | Atom |
cType | LV2_Atom_Bool |
owl:onDatatype | xsd:boolean |
An Int where 0 is false and any other value is true.
Chunk
ClassLabel | Chunk |
---|---|
Subclass of | Atom |
owl:onDatatype | xsd:base64Binary |
A chunk of memory with undefined contents.
This type is used to indicate a certain amount of space is available. For example, output ports with a variably sized type are connected to a Chunk so the plugin knows the size of the buffer available for writing.
The use of a Chunk should be constrained to a local scope, since interpreting it is impossible without context. However, if serialised to RDF, a Chunk may be represented directly as an xsd:base64Binary string, for example:
[] eg:someChunk "vu/erQ=="^^xsd:base64Binary .
Double
ClassLabel | Double |
---|---|
Subclass of | Number |
cType | LV2_Atom_Double |
owl:onDatatype | xsd:double |
A native `double`.
Event
ClassLabel | Event |
---|---|
cType | LV2_Atom_Event |
An atom with a time stamp prefix in a sequence.
An Event is typically an element of an Sequence. Note that this is not an Atom type since it begins with a timestamp, not an atom header.
Float
ClassLabel | Float |
---|---|
Subclass of | Number |
cType | LV2_Atom_Float |
owl:onDatatype | xsd:float |
A native `float`.
Literal
ClassLabel | Literal |
---|---|
Subclass of | Atom |
cType | LV2_Atom_Literal |
A UTF-8 string literal with optional datatype or language.
This type is compatible with rdfs:Literal and is capable of expressing a
string in any language or a value of any type. A Literal has a
datatype
and lang
followed by string data in UTF-8
encoding. The length of the string data in bytes is size -
sizeof(LV2_Atom_Literal)
, including the terminating NULL character. The
lang
field SHOULD be a URI of the form
http://lexvo.org/id/iso639-3/LANG
or
http://lexvo.org/id/iso639-1/LANG
where LANG is a 3-character ISO 693-3
language code, or a 2-character ISO 693-1 language code, respectively.
A Literal may have a datatype
or a lang
, but never both.
For example, a Literal can be Hello
in English:
void set_to_hello_in_english(LV2_Atom_Literal* lit) {
lit->atom.type = map(expand("Literal"));
lit->atom.size = 14;
lit->body.datatype = 0;
lit->body.lang = map("http://lexvo.org/id/iso639-1/en");
memcpy(LV2_ATOM_CONTENTS(LV2_Atom_Literal, lit),
"Hello",
sizeof("Hello")); // Assumes enough space
}
or a Turtle string:
void set_to_turtle_string(LV2_Atom_Literal* lit, const char* ttl) {
lit->atom.type = map(expand("Literal"));
lit->atom.size = 64;
lit->body.datatype = map("http://www.w3.org/2008/turtle#turtle");
lit->body.lang = 0;
memcpy(LV2_ATOM_CONTENTS(LV2_Atom_Literal, lit),
ttl,
strlen(ttl) + 1); // Assumes enough space
}
Long
ClassLabel | Long |
---|---|
Subclass of | Number |
cType | LV2_Atom_Long |
owl:onDatatype | xsd:long |
A native `int64_t`.
Number
ClassObject
ClassLabel | Object |
---|---|
Subclass of | Atom |
Superclass of | Blank |
Resource | |
cType | LV2_Atom_Object |
A collection of properties.
An Object
is an atom with a set of properties. This corresponds to an
RDF Resource, and can be thought of as a dictionary with URID keys.
An LV2_Atom_Object body has a uint32_t id
and type
, followed by a series of
Property bodies (LV2_Atom_Property_Body). The LV2_Atom_Object_Body::otype
field is equivalent to a property with key rdf:type, but is included in the
structure to allow for fast dispatching.
Code SHOULD check for objects using lv2_atom_forge_is_object() or lv2_atom_forge_is_blank() if a forge is available, rather than checking the atom type directly. This will correctly handle the deprecated Resource and Blank types.
When serialised to RDF, an Object is represented as a resource, for example:
eg:someObject
eg:firstPropertyKey "first property value" ;
eg:secondPropertyKey "first loser" ;
eg:andSoOn "and so on" .
Path
ClassLabel | Path |
---|---|
Subclass of | URI |
owl:onDatatype | URI |
A local file path.
A Path is a URI reference with only a path component: no scheme, authority, query, or fragment. In particular, paths to files in the same bundle may be cleanly written in Turtle files as a relative URI. However, implementations may assume any binary Path (e.g. in an event payload) is a valid file path which can passed to system functions like fopen() directly, without any character encoding or escape expansion required.
Any implementation that creates a Path atom to transmit to another is responsible for ensuring it is valid. A Path SHOULD always be absolute, unless there is some mechanism in place that defines a base path. Since this is not the case for plugin instances, effectively any Path sent to or received from a plugin instance MUST be absolute.
Property
ClassLabel | Property |
---|---|
Subclass of | Atom |
cType | LV2_Atom_Property |
A property of an Object.
An LV2_Atom_Property has a URID key
and context
, and an Atom value
. This
corresponds to an RDF Property, where the key
is the predicate
and the value
is the object.
The context
field can be used to specify a different context for each
property, where this is useful. Otherwise, it may be 0.
Properties generally only exist as part of an Object. Accordingly, they will typically be represented directly as properties in RDF (see Object). If this is not possible, they may be expressed as partial reified statements, for example:
eg:someProperty
rdf:predicate eg:theKey ;
rdf:object eg:theValue .
Sequence
ClassLabel | Sequence |
---|---|
Subclass of | Atom |
cType | LV2_Atom_Sequence |
A sequence of events.
A flat sequence of Event, that is, a series of time-stamped Atoms.
LV2_Atom_Sequence_Body.unit describes the time unit for the contained atoms. If the unit is known from context (e.g. run() stamps are always audio frames), this field may be zero. Otherwise, it SHOULD be either units:frame or units:beat, in which case ev.time.frames or ev.time.beats is valid, respectively.
If serialised to RDF, a Sequence has a similar form to Vector, but for brevity the elements may be assumed to be Event, for example:
eg:someSequence
a atom:Sequence ;
rdf:value (
[
atom:frameTime 1 ;
rdf:value "901A01"^^midi:MidiEvent
] [
atom:frameTime 3 ;
rdf:value "902B02"^^midi:MidiEvent
]
) .
Sound
ClassLabel | Sound |
---|---|
Subclass of | Vector |
cType | LV2_Atom_Vector |
A Vector of Float which represents an audio waveform.
The format of a Sound is the same as the buffer format for lv2:AudioPort (except the size may be arbitrary). An Sound inherently depends on the sample rate, which is assumed to be known from context. Because of this, directly serialising an Sound is probably a bad idea, use a standard format like WAV instead.
String
ClassLabel | String |
---|---|
Subclass of | Atom |
Superclass of | URI |
cType | LV2_Atom_String |
owl:onDatatype | xsd:string |
A UTF-8 string.
The body of an LV2_Atom_String is a C string in UTF-8 encoding, i.e. an array
of bytes (uint8_t
) terminated with a NULL byte ('\0'
).
This type is for free-form strings, but SHOULD NOT be used for typed data or text in any language. Use Literal unless translating the string does not make sense and the string has no meaningful datatype.
Tuple
ClassLabel | Tuple |
---|---|
Subclass of | Atom |
A sequence of atoms with varying type and size.
The body of a Tuple is simply a series of complete atoms, each aligned to 64 bits.
If serialised to RDF, a Tuple SHOULD have the form:
eg:someVector
a atom:Tuple ;
rdf:value (
"1"^^xsd:int
"3.5"^^xsd:float
"etc"
) .
URI
ClassLabel | URI |
---|---|
Subclass of | String |
Superclass of | Path |
owl:onDatatype | xsd:anyURI |
A URI string.
This is useful when a URI is needed but mapping is inappropriate, for example with temporary or relative URIs. Since the ability to distinguish URIs from plain strings is often necessary, URIs MUST NOT be transmitted as String.
This is not strictly a URI, since UTF-8 is allowed. Escaping and related issues are the host's responsibility.
URID
ClassLabel | URID |
---|---|
Subclass of | Atom |
cType | LV2_Atom_URID |
An unsigned 32-bit integer ID for a URI.
A URID is typically generated with the LV2_URID_Map provided by the host .
Vector
ClassLabel | Vector |
---|---|
Subclass of | Atom |
Superclass of | Sound |
cType | LV2_Atom_Vector |
A homogeneous sequence of atom bodies with equivalent type and size.
A homogeneous series of atom bodies with equivalent type and size.
An LV2_Atom_Vector is a 32-bit child_size
and child_type
followed by size
/ child_size
atom bodies.
For example, an Vector containing 42 elements of type Float:
struct VectorOf42Floats {
uint32_t size; // sizeof(LV2_Atom_Vector_Body) + (42 * sizeof(float);
uint32_t type; // map(expand("Vector"))
uint32_t child_size; // sizeof(float)
uint32_t child_type; // map(expand("Float"))
float elems[42];
};
Note that it is possible to construct a valid Atom for each element of the
vector, even by an implementation which does not understand child_type
.
If serialised to RDF, a Vector SHOULD have the form:
eg:someVector
a atom:Vector ;
atom:childType atom:Int ;
rdf:value (
"1"^^xsd:int
"2"^^xsd:int
"3"^^xsd:int
"4"^^xsd:int
) .
Properties
bufferType
PropertyLabel | buffer type |
---|---|
Domain | AtomPort |
Range | rdfs:Class |
Type | Object Property |
An atom type that a port may be connected to.
Indicates that an AtomPort may be connected to a certain Atom type. A port MAY support several buffer types. The host MUST NOT connect a port to an Atom with a type not explicitly listed with this property. The value of this property MUST be a sub-class of Atom. For example, an input port that is connected directly to an LV2_Atom_Double value is described like so:
<plugin>
lv2:port [
a lv2:InputPort , atom:AtomPort ;
atom:bufferType atom:Double ;
] .
This property only describes the types a port may be directly connected to. It says nothing about the expected contents of containers. For that, use supports.
cType
PropertyLabel | C type |
---|---|
Domain | rdfs:Class |
Range | lv2:Symbol |
Type | Datatype Property |
The C type that describes the binary representation of an Atom type.
frameTime
PropertyLabel | frame time |
---|---|
Range | xsd:decimal |
Type | Datatype Property |
A time stamp in audio frames.
supports
PropertyLabel | supports |
---|---|
Range | rdfs:Class |
Type | Object Property |
A supported atom type.
This property is defined loosely, it may be used to indicate that anything
supports
an Atom type, wherever that may be useful. It applies
recursively
where collections are involved.
In particular, this property can be used to describe which event types are expected by a port. For example, a port that receives MIDI events is described like so:
<plugin>
lv2:port [
a lv2:InputPort , atom:AtomPort ;
atom:bufferType atom:Sequence ;
atom:supports midi:MidiEvent ;
] .
Instances
atomTransfer
InstanceLabel | atom transfer |
---|---|
Type | ui:PortProtocol |
A port protocol for transferring atoms.
Transfer of the complete atom in a port buffer. Useful as the format
for a
LV2UI_Write_Function.
This protocol applies to atom ports. The host must transfer the complete atom contained in the port, including header.
eventTransfer
InstanceLabel | event transfer |
---|---|
Type | ui:PortProtocol |
A port protocol for transferring events.
Transfer of individual events in a port buffer. Useful as the format
for a
LV2UI_Write_Function.
This protocol applies to ports which contain events, usually in an Sequence. The host must transfer each individual event to the recipient. The format of the received data is an LV2_Atom, there is no timestamp header.
Abstract base class for all atoms.
An LV2_Atom has a 32-bit
size
andtype
, followed by a body ofsize
bytes. Atoms MUST be 64-bit aligned.All concrete Atom types (subclasses of this class) MUST define a precise binary layout for their body.
The
type
field is the URI of an Atom type mapped to an integer. Implementations SHOULD gracefully pass through, or ignore, atoms with unknown types.All atoms are POD by definition except references, which as a special case have
type
0. An Atom MUST NOT contain a Reference. It is safe to copy any non-reference Atom with a simplememcpy
, even if the implementation does not understandtype
. Though this extension reserves the type 0 for references, the details of reference handling are currently unspecified. A future revision of this extension, or a different extension, may define how to use non-POD data and references. Implementations MUST NOT send references to another implementation unless the receiver is explicitly known to support references (e.g. by supporting a feature).The special case of a null atom with both
type
andsize
0 is not considered a reference.