fix(astBuilder): fix export statment

[yjhmelody]

Now export * from "xx" will be built to export {} from "xx".

@willemneal visitor-as also has this problem.

It makes the behavior of the compiler plug-in difficult to understand, because there may be many files that are modified.

[yjhmelody]

@willemneal Can you take a look?

[willemneal]

I'm still unsure what the issue is here. Is it that an AST from export * from "" when turned back into a string became export {} from "" and now it will stay the same?

[yjhmelody]

I'm still unsure what the issue is here. Is it that an AST from export * from "" when turned back into a string became export {} from "" and now it will stay the same?

Yes, ASTbuilder does not handle the situation when the members of ExportStatement are null.

In program.ts We can see that when members is null, it means the ‘*' syntax.

  /** Initializes an `export` statement. */
  private initializeExports(
    /** The statement to initialize. */
    statement: ExportStatement,
    /** Parent file. */
    parent: File,
    /** So far queued `export`s. */
    queuedExports: Map<File,Map<string,QueuedExport>>,
    /** So far queued `export *`s. */
    queuedExportsStar: Map<File,QueuedExportStar[]>
  ): void {
    var members = statement.members;
    if (members) { // export { foo, bar } [from "./baz"]
      for (let i = 0, k = members.length; i < k; ++i) {
        this.initializeExport(members[i], parent, statement.internalPath, queuedExports);
      }
    } else { // export * from "./baz"
      let queued: QueuedExportStar[];
      if (queuedExportsStar.has(parent)) queued = assert(queuedExportsStar.get(parent));
      else queuedExportsStar.set(parent, queued = []);
      let foreignPath = statement.internalPath!; // must be set for export *
      queued.push(new QueuedExportStar(
        foreignPath,
        foreignPath.endsWith(INDEX_SUFFIX) // strip or add index depending on what's already present
          ? foreignPath.substring(0, foreignPath.length - INDEX_SUFFIX.length)
          : foreignPath + INDEX_SUFFIX,
        assert(statement.path)
      ));
    }
  }

[willemneal]

Okay makes sense to me.

[yjhmelody]

@MaxGraey Hi, anything wrong ?