Since it is essentially allocating non-gc managed objects,
it can lead to memory leaks and shouldn't happen invisibly.
Also support using the 'assign' key to request assign()
from another object before processing the current map.
E.g. df.global.cursor = { x = 1, y = 2, z = 3 }. The lua data
must be represented by raw lua tables.
For structs, the entries in the table are assigned to matching fields.
For containers, if a 'resize' field is missing or nil, the table is
treated like 1-based lua array, and the container is resized to match
its # length. Otherwise, the field must be either an explicit number,
true or false. If it is true, the size is selected by the highest index
in the table. After that, entries are copied using 0-based indices.
For pointers, the table must match the target object. If the pointer
is null, the object is auto-allocated; this can be controlled using
the 'new' field, the value of which will be passed to df.new().
Since it is essentially allocating non-gc managed objects,
it can lead to memory leaks and shouldn't happen invisibly.
Also support using the 'assign' key to request assign()
from another object before processing the current map.
E.g. df.global.cursor = { x = 1, y = 2, z = 3 }. The lua data
must be represented by raw lua tables.
For structs, the entries in the table are assigned to matching fields.
For containers, if a 'resize' field is missing or nil, the table is
treated like 1-based lua array, and the container is resized to match
its # length. Otherwise, the field must be either an explicit number,
true or false. If it is true, the size is selected by the highest index
in the table. After that, entries are copied using 0-based indices.
For pointers, the table must match the target object. If the pointer
is null, the object is auto-allocated; this can be controlled using
the 'new' field, the value of which will be passed to df.new().