build type based on provider, detect properly

src/universal_lambda_build.zig

@@ -1,46 +1,62 @@
 const std = @import("std");
+
+/// Determines the style of interface between what is in a main program
+/// and the provider system. This should not be an exhaustive set of steps,
+/// but a higher level "how do I get the handler registered". So only what
+/// would describe a runtime difference
 pub const BuildType = enum {
-    awslambda_package,
-    awslambda_iam,
-    awslambda_deploy,
-    awslambda_run,
+    awslambda,
     exe_run,
     standalone_run,
-    // cloudflare_* (TBD)
-    // flexilib_* (TBD)
+    cloudflare,
+    flexilib,
 };
-// awslambda_package
-// awslambda_iam
-// awslambda_deploy
-// awslambda_run
-// exe_run // TODO: Can we skip this?
-// cloudflare_* (TBD)
-// flexilib_* (TBD)
-pub fn configureBuild(b: *std.Build, exe: *std.Build.Step.Compile) !void {
-    // Make our target platform visible to runtime through an import
-    // called "build_options"
-    var options_module: *std.Build.Module = undefined;
-    {
-        // We need to go through the command line args, look for argument(s)
-        // between "build" and anything prefixed with "-". First take, blow up
-        // if there is more than one. That's the step we're rolling with
-        // These frameworks I believe are inextricably tied to both build and
-        // run behavior
-        const options = b.addOptions();
-        options.addOption(BuildType, "build_type", .exe_run);
-        exe.addOptions("build_options", options);
-        options_module = exe.modules.get("build_options").?;
-    }
-    // Add modules
-    {
-        exe.addAnonymousModule("universal_lambda_handler", .{
-            .source_file = .{ .path = "upstream/src/universal_lambda.zig" },
-            .dependencies = &[_]std.Build.ModuleDependency{.{
-                .name = "build_options",
-                .module = options_module,
-            }},
-        });
-    }
 
+pub fn configureBuild(b: *std.Build, exe: *std.Build.Step.Compile) !void {
     // Add steps
+    try @import("lambdabuild.zig").configureBuild(b, exe);
+    // Add options module so we can let our universal_lambda know what
+    // type of interface is necessary
+
+    // Add module
+    exe.addAnonymousModule("universal_lambda_handler", .{
+        .source_file = .{ .path = "upstream/src/universal_lambda.zig" },
+        .dependencies = &[_]std.Build.ModuleDependency{.{
+            .name = "build_options",
+            .module = try createOptionsModule(b, exe),
+        }},
+    });
+}
+
+/// Make our target platform visible to runtime through an import
+/// called "build_options". This will also be available to the consuming
+/// executable if needed
+fn createOptionsModule(b: *std.Build, exe: *std.Build.Step.Compile) !*std.Build.Module {
+    // We need to go through the command line args, look for argument(s)
+    // between "build" and anything prefixed with "-". First take, blow up
+    // if there is more than one. That's the step we're rolling with
+    // These frameworks I believe are inextricably tied to both build and
+    // run behavior.
+    //
+    var args = try std.process.argsAlloc(b.allocator);
+    defer b.allocator.free(args);
+    const options = b.addOptions();
+    options.addOption(BuildType, "build_type", findBuildType(args) orelse .exe_run);
+    exe.addOptions("build_options", options);
+    return exe.modules.get("build_options").?;
+}
+
+fn findBuildType(build_args: [][:0]u8) ?BuildType {
+    var rc: ?BuildType = null;
+    for (build_args[1..]) |arg| {
+        if (std.mem.startsWith(u8, arg, "-")) break; // we're done as soon as we get to options
+        inline for (std.meta.fields(BuildType)) |field| {
+            if (std.mem.startsWith(u8, arg, field.name)) {
+                if (rc != null)
+                    @panic("Sorry, we are not smart enough to build a single handler for multiple simultaneous providers");
+                rc = @field(BuildType, field.name);
+            }
+        }
+    }
+    return rc;
 }