Handle each kind of module field separately in preprocessor #5

This does mean we have to temporarily match against the entire S-expression instead of peeking at it, but that’ll go away once we start consuming it incrementally. For now the priority is to get #read_list lifted out of #process_field so that it’s operating entirely inside a list like all the other methods do.

We were only wrapping `field` in a single-element array because it was being passed to the #read_list helper inside #process_field, but now the helper’s moved next to the array, the two cancel each other out.

Although this is slightly less efficient, it more directly resembles the specification of export abbreviations (e.g. [0]) and so should be easier to understand at a glance. [0] https://webassembly.github.io/spec/core/text/modules.html#text-global-abbrev

Just as in #expand_inline_export, we can use #process_command to desugar the result of expanding the module abbreviation [0]. [0] https://webassembly.github.io/spec/core/text/modules.html#id10

This makes it possible for each method to expand inline exports, which produce multiple fields. In practice it’s only #process_function that needs to do this, but I’ve made the same change to #process_type and #process_import to keep all of the field-processing methods consistent.

The spec calls these fields “function definitions” [0] and “type definitions” [1], so we should use the same terminology here. [0] https://webassembly.github.io/spec/core/text/modules.html#functions [1] https://webassembly.github.io/spec/core/text/modules.html#types

This removes the need to support inline imports & exports for functions in #process_field.

It’s handled in #process_function_definition now.

…itions This removes the need to support inline imports & exports for any kind of field in #process_field.

It’s entirely handled in field-specific methods now.

They’re called by both #process_function_definition and #process_table_memory_global_definition, and the code is easier to read if callers come before callees.

…xport We really want to get the optional ID out of the way, but we have to read the field kind first.

…rocess_table_memory_global_definition

We need to do the same work to handle inline imports & exports for function and table/memory/global definitions, so it’s convenient to move the duplicated code into helpers.

We’d like to lift the #read out of this method, so we can’t use it directly in a conditional. Fortunately we can check its side-effect (of assigning to `id`) instead.

Rather than doing the reads inline when they’re needed, we read everything together first and then build the result. The increased similarity to #expand_inline_import makes the methods easier to compare at a glance.

…able_memory_global_definition

We now only need to look at the next thing we’ll read, not the entire S-expression.

Table, memory and global definitions support different abbreviations, so each will need its own slightly different method.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Handle each kind of module field separately in preprocessor #5

Handle each kind of module field separately in preprocessor #5

Commits on Jun 11, 2023