refactor: decompose macro parser into dedicated parsing stages #3

watta0 · 2026-05-30T17:32:39Z

watta0 commented

2026-05-30 17:32:39 +00:00

Summary

Refactor the macro parser by decomposing the monolithic parsing logic into dedicated parsing stages and introducing a clearer state management model.

This change significantly improves readability, maintainability, and future extensibility while preserving the original behavior of the parser.

Motivation

The previous implementation contained most parsing logic inside main() and relied heavily on labels and cross-scope jumps. While functional, the control flow was difficult to follow and made future modifications risky.

This refactor aims to:

Reduce complexity in main()
Isolate parsing responsibilities
Improve naming consistency
Centralize error handling
Make individual parsing stages easier to understand and test

Changes

Parser decomposition

Split the parser into dedicated functions:

s_parse_book()
- Detects and validates the ae2f_MAC( sequence.
s_parse_tparam()
- Parses template parameter declarations.
s_parse_fnname()
- Parses and emits function/named block.
s_parse_params()
- Parses macro parameters and generates parameter comments.
s_parse_body()
- Parses macro bodies and handles newline continuation conversion.
s_parse_include()
- Handles #include directives.
s_parse_string()
- Handles string literals and escape sequences.
s_parse_else()
- Handles ordinary character output.

State management cleanup

Replaced anonymous state values with explicit enums:

enum RET_STATE_
enum FLAG_

This makes parser states and error conditions self-documenting and removes magic numbers from the implementation.

Naming improvements

Renamed several macros and constants to follow a more descriptive naming convention.

Examples:

SZPARAM → PARAM_LEN
SZTPARAM → TPARAM_LEN
isAlph() → IS_ALPHA()
isNumber() → IS_NUMBER()
isVarName() → IS_VARNAME()
isNewLine() → IS_NEWLINE()

Flag operations were also renamed:

isTrue() → COMPARE_FLAG()
setTrue() → SET_FLAG_TRUE()
setFalse() → SET_FLAG_FALSE()

Unified error handling

Introduced helper macros:

__RET_IF_FAILED()
__RET_IF_EOF()

Benefits:

Eliminates repetitive error-checking code.
Makes failure paths consistent.
Improves readability of parser logic.

Buffer and constant cleanup

Introduced clearer length-related constants:

BOOK_LEN
PARAM_LEN
TPARAM_LEN

and removed several ad-hoc size calculations from parsing logic.

Improved separation of concerns

main() now acts primarily as a dispatcher:

Detect token category
Invoke the appropriate parser
Propagate parser results

This significantly reduces the amount of logic embedded in the main execution loop.

Benefits

Easier to understand parser flow
Smaller and more focused functions
Reduced cognitive complexity
Improved maintainability
Easier future feature additions
More consistent error handling
Better testability of individual parser stages

Notes

This refactor is intended to be behavior-preserving. No functional changes to the generated macro output are expected.

## Summary Refactor the macro parser by decomposing the monolithic parsing logic into dedicated parsing stages and introducing a clearer state management model. This change significantly improves readability, maintainability, and future extensibility while preserving the original behavior of the parser. ## Motivation The previous implementation contained most parsing logic inside `main()` and relied heavily on labels and cross-scope jumps. While functional, the control flow was difficult to follow and made future modifications risky. This refactor aims to: * Reduce complexity in `main()` * Isolate parsing responsibilities * Improve naming consistency * Centralize error handling * Make individual parsing stages easier to understand and test ## Changes ### Parser decomposition Split the parser into dedicated functions: * `s_parse_book()` * Detects and validates the `ae2f_MAC(` sequence. * `s_parse_tparam()` * Parses template parameter declarations. * `s_parse_fnname()` * Parses and emits function/named block. * `s_parse_params()` * Parses macro parameters and generates parameter comments. * `s_parse_body()` * Parses macro bodies and handles newline continuation conversion. * `s_parse_include()` * Handles `#include` directives. * `s_parse_string()` * Handles string literals and escape sequences. * `s_parse_else()` * Handles ordinary character output. ### State management cleanup Replaced anonymous state values with explicit enums: ```c enum RET_STATE_ enum FLAG_ ``` This makes parser states and error conditions self-documenting and removes magic numbers from the implementation. ### Naming improvements Renamed several macros and constants to follow a more descriptive naming convention. Examples: * `SZPARAM` → `PARAM_LEN` * `SZTPARAM` → `TPARAM_LEN` * `isAlph()` → `IS_ALPHA()` * `isNumber()` → `IS_NUMBER()` * `isVarName()` → `IS_VARNAME()` * `isNewLine()` → `IS_NEWLINE()` Flag operations were also renamed: * `isTrue()` → `COMPARE_FLAG()` * `setTrue()` → `SET_FLAG_TRUE()` * `setFalse()` → `SET_FLAG_FALSE()` ### Unified error handling Introduced helper macros: ```c __RET_IF_FAILED() __RET_IF_EOF() ``` Benefits: * Eliminates repetitive error-checking code. * Makes failure paths consistent. * Improves readability of parser logic. ### Buffer and constant cleanup Introduced clearer length-related constants: ```c BOOK_LEN PARAM_LEN TPARAM_LEN ``` and removed several ad-hoc size calculations from parsing logic. ### Improved separation of concerns `main()` now acts primarily as a dispatcher: * Detect token category * Invoke the appropriate parser * Propagate parser results This significantly reduces the amount of logic embedded in the main execution loop. ## Benefits * Easier to understand parser flow * Smaller and more focused functions * Reduced cognitive complexity * Improved maintainability * Easier future feature additions * More consistent error handling * Better testability of individual parser stages ## Notes This refactor is intended to be behavior-preserving. No functional changes to the generated macro output are expected.

watta0 added 2 commits

2026-05-30 17:32:39 +00:00

tmp d9abef3854

FIX: massive refactor a884189237

dalmurii requested changes

2026-05-30 18:14:36 +00:00

Macro/main.c

					
				@ -332,4 +334,0 @@

										return 1;

				casenewlines:

										ENDL[2] = (char) ch;

ENDL is used only here and modified. Declare this variable as under s_parse_body.

ENDL is used only here and modified. Declare this variable as under `s_parse_body`.

got it!

Macro/main.c Outdated

					
				@ -1,22 +1,21 @@

				#define	_CRT_SECURE_NO_WARNINGS	1

				#include <.main.h>

				#include <assert.h>

That header is used nowhere.

Order received!

Macro/main.c Outdated

					
				@ -18,2 +16,2 @@

				#define BOOK (ae2f_MAC_KEYWORD "(")

				#define	STRLEN(a)	((int)(sizeof(a) - 1))

				#define BOOK		(ae2f_MAC_KEYWORD "(")

				#define BOOK_LEN		((size_t)(sizeof(BOOK) - (size_t)1))

Duplicated. Would you utilise STRLEN to make BOOK_LEN?

Got it!

Macro/main.c

					
				@ -66,0 +46,4 @@

				#define	__RET_IF_FAILED(ret_var, expr)		\

					do {					\

						(ret_var) = (expr);		\

						if ((ret_var) != RET_STATE_OK)	\

same as #3/files (comment)

same as https://talkischeap.work.gd/ae2f/Preproc/pulls/3/files#issuecomment-154

Got it

Macro/main.c

					
				@ -66,0 +53,4 @@

				#define	__RET_IF_EOF(expr)				\

					do {						\

						if ((expr) == EOF)			\

							return RET_STATE_OUTFAILED;	\

enum definition is below this macro. move that please.

Got it

Macro/main.c Outdated

					
				@ -66,0 +80,4 @@

				static enum RET_STATE_ s_parse_fnname (flags_t *rdwr_flags);

				static enum RET_STATE_ s_parse_params (flags_t *rdwr_flags);

				static enum RET_STATE_ s_parse_body   (flags_t *rdwr_flags);

				static enum RET_STATE_ s_parse_else   (int ch);

This name does not explain what is this else of.
I suggest s_parse_fallback of s_parse_default

This name does not explain what is this else of. I suggest `s_parse_fallback` of `s_parse_default`

Got it

Macro/main.c

					
				@ -94,1 +99,3 @@

								return STATE_OUTFAILED;

						switch (ch) {

						case '\"':

							__RET_IF_FAILED(ret, s_parse_string(ch));

You know what ch is. Using literal would be helpful.

You know what `ch` is. Using literal would be helpful.

Order received

Macro/main.c

					
				@ -95,0 +101,4 @@

							__RET_IF_FAILED(ret, s_parse_string(ch));

							break;

						case '#':

							__RET_IF_FAILED(ret, s_parse_include(ch));

Same as #3/files (comment)

Same as https://talkischeap.work.gd/ae2f/Preproc/pulls/3/files#issuecomment-156

Got it

Macro/main.c Outdated

					
				@ -289,2 +308,2 @@

										PARAMS_BUF[prms_len] = (char) ch;

										PARAMS_BUF[prms_len + 1] = 0;

							buf[buf_len] = (char)ch;

							buf[buf_len + 1] = '\0';

That needs former length test

got it

Macro/main.c

					
				@ -303,2 +315,2 @@

											PARAMS_BUF[prms_len + 4] = 0;

										} 

							if (ch == ')') {

								buf[buf_len]     = ' ';

That is unnecessary as I see now.
This was my mistake and you seem to be shipping it as is.

Putting putsas is and printing this statement later would be more stable.

That is unnecessary as I see now. This was my mistake and you seem to be shipping it as is. Putting `puts`as is and printing this statement later would be more stable.