poly-gui / nanoc Goto Github PK

Currently, a field number must be explicitly assigned for every field in a message. The field numbers only indicate the order of the fields in the serialized bytes. For example, field number 0 appears first in the buffer, then field number 1, so on and so forth. Field numbers are needed because the YAML parser doesn't guarantee that the fields are traversed in the order they appear in the schema file, and the numbers are therefore used to give the compiler a consistent order across invocations.

In a7609e1, the YAML parser was downgraded to introduce the MapSlice struct that preserves the order of the fields as they appear in the schema file. This eliminates the need for explicit field numbers, which reduces noise in the schema file.

Currently, a type ID must be explicitly specified for every NanoPack message in its schema. As the project grows, this will be completely unmaintainable. To mitigate this, the compiler should be able to:

• automatically assign a new type ID for every new message defined; and
• persist the assigned type IDs so that they are consistent for every compiler invocation.

To do so, the compiler should follow the following algorithm:

check whether the generated code for a schema exists
if it exists:
    read code header and check if a type ID is assigned
    if not exist:
        generate a new type ID
        write the type ID to the code header
else:
    generate a new type ID
    write the type ID to the code header