From 5da855c331d20614c004f366ffcb571d9e6861c7 Mon Sep 17 00:00:00 2001
From: John Kerl <kerl.john.r@gmail.com>
Date: Sat, 22 Aug 2015 21:08:49 -0400
Subject: [PATCH] [csv iterate] string-builder iterate

---
 c/csv-rfc.txt          |  9 +++++++++
 c/lib/string_builder.c | 19 ++++++++++---------
 2 files changed, 19 insertions(+), 9 deletions(-)

diff --git a/c/csv-rfc.txt b/c/csv-rfc.txt
index dc546678c..ba20be10c 100644
--- a/c/csv-rfc.txt
+++ b/c/csv-rfc.txt
@@ -1,6 +1,13 @@
+
 ================================================================
 --> parser option: write a grammar like the below. be careful to
     preserve the streaming property!
+--> I don't see a way to line-at-a-time it in the formal grammar.
+    Other option is to find end of line via string arithmetic,
+    then pass that (null-terminated) string to the parser. But
+    correctly determining EOL in the presence of double-quote
+    semantics *is* the problem; nothing interesting left for the
+    parser to do.
 
 --> state-machine option:
     * each field either does or does not begin & end w/ "
@@ -24,6 +31,8 @@
     public fcn append_char, public fcn append_string,
     private fcn realloc resizing with minimal copy, public fcn zterm & produce.
 
+================================================================
+Offlineable (subway-tunnelable) excerpt from https://tools.ietf.org/html/rfc4180
 ================================================================
 Definition of the CSV Format
 
diff --git a/c/lib/string_builder.c b/c/lib/string_builder.c
index 099c653c0..c7846237b 100644
--- a/c/lib/string_builder.c
+++ b/c/lib/string_builder.c
@@ -4,14 +4,10 @@
 #include "../lib/mlrutil.h"
 #include "../lib/mlr_globals.h"
 
+// Private method:
 static void sb_enlarge(string_builder_t* psb);
 
-// typedef struct _string_builder_t {
-//         int used_length;
-//         int alloc_length;
-//         char* buffer;
-// } string_builder_t;
-
+// ----------------------------------------------------------------
 void sb_init(string_builder_t* psb, int alloc_length) {
 	if (alloc_length < 1) {
 		fprintf(stderr, "%s: string_builder alloc_length must be >= 1; got %d.\n",
@@ -23,26 +19,31 @@ void sb_init(string_builder_t* psb, int alloc_length) {
 	psb->buffer = mlr_malloc_or_die(alloc_length); // xxx malloc ...
 }
 
+// ----------------------------------------------------------------
 void sb_append_char(string_builder_t* psb, char c) {
 	if (psb->used_length >= psb->alloc_length)
 		sb_enlarge(psb);
 	psb->buffer[psb->used_length++] = c;
 }
 
+// ----------------------------------------------------------------
 void sb_append_string(string_builder_t* psb, char* s) {
 	for (char* p = s; *p; p++)
 		sb_append_char(psb, *p);
 }
 
+// ----------------------------------------------------------------
+// Keep and reuse the internal buffer. Allocate to the caller only
+// the size needed.
 char* sb_finish(string_builder_t* psb) {
 	sb_append_char(psb, '\0');
-	char* rv = psb->buffer;
+	char* rv = mlr_malloc_or_die(psb->used_length);
+	memcpy(rv, psb->buffer, psb->used_length);
 	psb->used_length  = 0;
-	psb->alloc_length = 0;
-	psb->buffer       = NULL;
 	return rv;
 }
 
+// ----------------------------------------------------------------
 static void sb_enlarge(string_builder_t* psb) {
 	int new_alloc_length = psb->alloc_length * 2;
 	char* new_buffer = mlr_malloc_or_die(new_alloc_length);