gbranden pushed a commit to branch master
in repository groff.
commit 034b847a2d19891c0bea5437815c02e863076b80
Author: G. Branden Robinson <g.branden.robin...@gmail.com>
AuthorDate: Thu Jul 10 20:39:03 2025 -0500
src/roff/troff/input.cpp: Refactor to simplify.
* src/roff/troff/input.cpp: As an apparent optimization (dating back as
far as groff 1.02, 1991) parts of the input parser that needed to read
arbitrarily far ahead in the input stream to consume a token used an
initial stack-allocated buffer of 16 bytes. (Also known as
"automatic" storage, thus the name "abuf", I think.) If 16 bytes
wasn't enough, open-coded logic would allocate a buffer from the heap
and double it in size until the token fit. Simplify by converting
`ABUF_SIZE` from a preprocessor macro to a `const` symbol named
`default_buffer_size`, and _always_ allocate from the heap. In the
future, we can convert stuff like this to a C++ STL `vector` and let
the standard library worry about dynamic storage management.
(read_long_escape_parameters, do_get_long_name, get_delimited_name):
Do it.
---
ChangeLog | 18 +++++++
src/roff/troff/input.cpp | 137 ++++++++++++++++-------------------------------
2 files changed, 63 insertions(+), 92 deletions(-)
diff --git a/ChangeLog b/ChangeLog
index 0b6b81fde..fd18e1893 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,21 @@
+2025-07-10 G. Branden Robinson <g.branden.robin...@gmail.com>
+
+ * src/roff/troff/input.cpp: Refactor to simplify. As an
+ apparent optimization (dating back as far as groff 1.02, 1991)
+ parts of the input parser that needed to read arbitrarily far
+ ahead in the input stream to consume a token used an initial
+ stack-allocated buffer of 16 bytes. (Also known as "automatic"
+ storage, thus the name "abuf", I think.) If 16 bytes wasn't
+ enough, open-coded logic would allocate a buffer from the heap
+ and double it in size until the token fit. Simplify by
+ converting `ABUF_SIZE` from a preprocessor macro to a `const`
+ symbol named `default_buffer_size`, and _always_ allocate from
+ the heap. In the future, we can convert stuff like this to a
+ C++ STL `vector` and let the standard library worry about
+ dynamic storage management.
+ (read_long_escape_parameters, do_get_long_name)
+ (get_delimited_name): Do it.
+
2025-07-10 G. Branden Robinson <g.branden.robin...@gmail.com>
* src/roff/troff/input.cpp: Fix code style nits.
diff --git a/src/roff/troff/input.cpp b/src/roff/troff/input.cpp
index b9acacd2c..4275e9f87 100644
--- a/src/roff/troff/input.cpp
+++ b/src/roff/troff/input.cpp
@@ -80,12 +80,12 @@ extern "C" {
(WARN_CHAR|WARN_NUMBER|WARN_BREAK|WARN_SPACE|WARN_FONT|WARN_FILE)
#endif
-// initial size of buffer for reading names; expanded as necessary
-#define ABUF_SIZE 16
-
extern "C" const char *program_name;
extern "C" const char *Version_string;
+// initial size for input buffers that need to grow arbitrarily
+static const int default_buffer_size = 16;
+
#ifdef COLUMN
void init_column_requests();
#endif /* COLUMN */
@@ -995,39 +995,28 @@ static symbol read_two_char_escape_parameter()
static symbol read_long_escape_parameters(read_mode mode)
{
int start_level = input_stack::get_level();
- char abuf[ABUF_SIZE];
- char *buf = abuf;
- int buf_size = ABUF_SIZE;
+ int buf_size = default_buffer_size;
+ char *buf = new char[buf_size];
int i = 0;
char c;
bool have_char = false;
for (;;) {
c = get_char_for_escape_parameter(have_char && (WITH_ARGS == mode));
if ('\0' == c) {
- if (buf != abuf)
- delete[] buf;
+ delete[] buf;
return NULL_SYMBOL;
}
have_char = true;
if ((WITH_ARGS == mode) && (' ' == c))
break;
- if ((i + 2) > buf_size) {
- if (buf == abuf) {
- // C++03: new char[ABUF_SIZE * 2]();
- buf = new char[ABUF_SIZE * 2];
- (void) memset(buf, 0, (ABUF_SIZE * 2 * sizeof(char)));
- memcpy(buf, abuf, buf_size);
- buf_size = ABUF_SIZE * 2;
- }
- else {
- char *old_buf = buf;
- // C++03: new char[buf_size * 2]();
- buf = new char[buf_size * 2];
- (void) memset(buf, 0, (buf_size * 2 * sizeof(char)));
- memcpy(buf, old_buf, buf_size);
- buf_size *= 2;
- delete[] old_buf;
- }
+ if (i + 2 > buf_size) {
+ char *old_buf = buf;
+ // C++03: new char[buf_size * 2]();
+ buf = new char[buf_size * 2];
+ (void) memset(buf, 0, (buf_size * 2 * sizeof(char)));
+ memcpy(buf, old_buf, buf_size);
+ buf_size *= 2;
+ delete[] old_buf;
}
if ((']' == c) && (input_stack::get_level() == start_level))
break;
@@ -1043,14 +1032,9 @@ static symbol read_long_escape_parameters(read_mode mode)
}
if (' ' == c)
have_multiple_params = true;
- if (buf == abuf) {
- return symbol(abuf);
- }
- else {
- symbol s(buf);
- delete[] buf;
- return s;
- }
+ symbol s(buf);
+ delete[] buf;
+ return s;
}
static symbol read_escape_parameter(read_mode mode)
@@ -2834,29 +2818,19 @@ static symbol do_get_long_name(bool required, char
end_char)
{
while (tok.is_space())
tok.next();
- char abuf[ABUF_SIZE];
- char *buf = abuf;
- int buf_size = ABUF_SIZE;
+ int buf_size = default_buffer_size;
+ char *buf = new char[buf_size];
int i = 0;
for (;;) {
// If `end_char` != `\0` we normally have to append a null byte.
if ((i + 2) > buf_size) {
- if (buf == abuf) {
- // C++03: new char[ABUF_SIZE * 2]();
- buf = new char[ABUF_SIZE * 2];
- (void) memset(buf, 0, (ABUF_SIZE * 2 * sizeof(char)));
- memcpy(buf, abuf, (buf_size * sizeof(char)));
- buf_size = ABUF_SIZE * 2;
- }
- else {
- char *old_buf = buf;
- // C++03: new char[buf_size * 2]();
- buf = new char[buf_size * 2];
- (void) memset(buf, 0, (buf_size * 2 * sizeof(char)));
- memcpy(buf, old_buf, (buf_size * sizeof(char)));
- buf_size *= 2;
- delete[] old_buf;
- }
+ char *old_buf = buf;
+ // C++03: new char[buf_size * 2]();
+ buf = new char[buf_size * 2];
+ (void) memset(buf, 0, (buf_size * 2 * sizeof(char)));
+ memcpy(buf, old_buf, (buf_size * sizeof(char)));
+ buf_size *= 2;
+ delete[] old_buf;
}
if ((buf[i] = tok.ch()) == '\0' || (buf[i] == end_char))
break;
@@ -2871,13 +2845,9 @@ static symbol do_get_long_name(bool required, char
end_char)
buf[i + 1] = '\0';
else
diagnose_invalid_identifier();
- if (buf == abuf)
- return symbol(buf);
- else {
- symbol s(buf);
- delete[] buf;
- return s;
- }
+ symbol s(buf);
+ delete[] buf;
+ return s;
}
static void close_all_streams();
@@ -5801,28 +5771,18 @@ static symbol get_delimited_name()
return NULL_SYMBOL;
}
int start_level = input_stack::get_level();
- char abuf[ABUF_SIZE];
- char *buf = abuf;
- int buf_size = ABUF_SIZE;
+ int buf_size = default_buffer_size;
+ char *buf = new char[buf_size];
int i = 0;
for (;;) {
if ((i + 1) > buf_size) {
- if (buf == abuf) {
- // C++03: new char[ABUF_SIZE * 2]();
- buf = new char[ABUF_SIZE * 2];
- (void) memset(buf, 0, (ABUF_SIZE * 2 * sizeof(char)));
- memcpy(buf, abuf, buf_size);
- buf_size = ABUF_SIZE * 2;
- }
- else {
- char *old_buf = buf;
- // C++03: new char[buf_size * 2]();
- buf = new char[buf_size * 2];
- (void) memset(buf, 0, (buf_size * 2 * sizeof(char)));
- memcpy(buf, old_buf, buf_size);
- buf_size *= 2;
- delete[] old_buf;
- }
+ char *old_buf = buf;
+ // C++03: new char[buf_size * 2]();
+ buf = new char[buf_size * 2];
+ (void) memset(buf, 0, (buf_size * 2 * sizeof(char)));
+ memcpy(buf, old_buf, buf_size);
+ buf_size *= 2;
+ delete[] old_buf;
}
tok.next();
if ((tok == start_token)
@@ -5838,26 +5798,19 @@ static symbol get_delimited_name()
error("closing delimiter does not match; expected %1, got %2",
delimdesc, tok.description());
free(delimdesc);
- if (buf != abuf)
- delete[] buf;
+ delete[] buf;
return NULL_SYMBOL;
}
i++;
}
buf[i] = '\0';
- if (buf == abuf) {
- if (0 == i) {
- error("empty delimited name");
- return NULL_SYMBOL;
- }
- else
- return symbol(buf);
- }
- else {
- symbol s(buf);
- delete[] buf;
- return s;
+ if (0 == i) {
+ error("empty delimited name");
+ return NULL_SYMBOL;
}
+ symbol s(buf);
+ delete[] buf;
+ return s;
}
static void do_register() // \R
_______________________________________________
groff-commit mailing list
groff-commit@gnu.org
https://lists.gnu.org/mailman/listinfo/groff-commit
