Metadata and Specs


Each Tiled node carries an optional dictionary of metadata. This is fully under the control of the user: Tiled itself is not opinionated about its content and does not reserve any names for its own use.


The metadata dictionary may be arbitrarily nested, and it may contain anything that Tiled can transmit as JSON or msgpack. All JSON-serializable types (dicts, lists, numbers, strings) are supported. The following types are additionally supported:

  • Dates, represented as datetime.datetime objects, are supported natively by msgpack. JSON does not support dates, so they are converted to ISO 8601 strings when clients request JSON.

  • Numpy objects are tolerated as input, but they are converted to (nested) lists before encoding to msgpack or JSON. Metadata is not a suitable place to store large arrays.

  • bytes objects are supported natively by msgpack. If JSON is requested, conversion to unicode is attempted.

  • uuid.UUID objects are converted into the standard hyphen-separated hex representation.


Every node in Tiled has exactly one “structure family” (“node”, “array”, “dataframe”, etc.) It’s useful to think of the structure family as a coarse, lowest-common-denominator description of how to query and interpret the data.

Sometimes, it is useful to be more specific than the structure family. Each Tiled node carries an optional list of “specs” (i.e. specifications). These are meant to communicate that the metadata and/or data conforms to some recognized layout, schema, or naming convention that may have meaning to clients. Clients that recognize a spec can use it to provide a fine-tuned user experience, such as more useful displays, specialized conveniences, and performance optimizations.

Specs are given as a list of strings, meant to be ordered from most specific to list specific. A spec may refer to a formally published specification or an ad hoc local convention. It is not necessary for every Tiled client to understand every spec in use. A client can walk the list of specs in order and stop if it finds a spec it recognizes. If a client does not recognize any of the specs in the list, or if no specs are given, it can fall back to the structure family to obtain a workable description of the data. Specs are just an upgrade: “If you know what this means, you can use it to assume additional constraints.”