aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorDavid Robillard <d@drobilla.net>2018-05-12 22:44:36 +0200
committerDavid Robillard <d@drobilla.net>2020-10-27 13:13:58 +0100
commit255fc4d36518728e8367feafb4a154e14b2b35d0 (patch)
tree4e5b6907ca9288ae2d3aa649cce7a39f83cf185a
parent95f7334bbc02f3e75b33b9914eee58c69f1588bd (diff)
downloadserd-255fc4d36518728e8367feafb4a154e14b2b35d0.tar.gz
serd-255fc4d36518728e8367feafb4a154e14b2b35d0.tar.bz2
serd-255fc4d36518728e8367feafb4a154e14b2b35d0.zip
Simplify stack management by popping in bulk at higher levels
Since all memory used by the reader is POD in the stack, there is no benefit to forcing code to explicitly pop everything pushed to the stack, since any function can record an offset and pop back down to it regardless of what its callers pushed if it knows that it does not need those items.
-rw-r--r--src/n3.c85
-rw-r--r--src/reader.c27
-rw-r--r--src/reader.h4
-rw-r--r--src/stack.h33
4 files changed, 51 insertions, 98 deletions
diff --git a/src/n3.c b/src/n3.c
index 190d7239..1b2c5a06 100644
--- a/src/n3.c
+++ b/src/n3.c
@@ -680,7 +680,6 @@ read_IRIREF(SerdReader* reader, SerdNode** dest)
*dest = push_node(reader, SERD_URI, "", 0);
if (!fancy_syntax(reader) && read_IRIREF_scheme(reader, *dest)) {
- *dest = pop_node(reader, *dest);
return r_err(reader, SERD_ERR_BAD_SYNTAX, "expected IRI scheme\n");
}
@@ -690,20 +689,17 @@ read_IRIREF(SerdReader* reader, SerdNode** dest)
const int c = eat_byte_safe(reader, peek_byte(reader));
switch (c) {
case '"': case '<': case '^': case '`': case '{': case '|': case '}':
- *dest = pop_node(reader, *dest);
return r_err(reader, SERD_ERR_BAD_SYNTAX,
"invalid IRI character `%c'\n", c);
case '>':
return SERD_SUCCESS;
case '\\':
if (read_UCHAR(reader, *dest, &code)) {
- *dest = pop_node(reader, *dest);
return r_err(reader, SERD_ERR_BAD_SYNTAX,
"invalid IRI escape\n");
}
switch (code) {
case 0: case ' ': case '<': case '>':
- *dest = pop_node(reader, *dest);
return r_err(reader, SERD_ERR_BAD_SYNTAX,
"invalid escaped IRI character U+%04X\n", code);
default:
@@ -722,7 +718,6 @@ read_IRIREF(SerdReader* reader, SerdNode** dest)
(unsigned)c);
}
if (reader->strict) {
- *dest = pop_node(reader, *dest);
return SERD_ERR_BAD_SYNTAX;
}
st = SERD_FAILURE;
@@ -731,14 +726,12 @@ read_IRIREF(SerdReader* reader, SerdNode** dest)
push_byte(reader, *dest, c);
} else if (read_utf8_character(reader, *dest, (uint8_t)c)) {
if (reader->strict) {
- *dest = pop_node(reader, *dest);
return SERD_ERR_BAD_SYNTAX;
}
}
}
}
- *dest = pop_node(reader, *dest);
return st;
}
@@ -871,7 +864,7 @@ read_literal(SerdReader* reader,
SerdStatus st = read_String(reader, *dest, flags);
if (st) {
- *dest = pop_node(reader, *dest);
+ *dest = NULL;
return st;
}
@@ -880,9 +873,6 @@ read_literal(SerdReader* reader,
eat_byte_safe(reader, '@');
*flags |= SERD_HAS_LANGUAGE;
if ((st = read_LANGTAG(reader, lang))) {
- *datatype = pop_node(reader, *datatype);
- *lang = pop_node(reader, *lang);
- *dest = pop_node(reader, *dest);
return r_err(reader, st, "bad literal\n");
}
break;
@@ -891,9 +881,6 @@ read_literal(SerdReader* reader,
eat_byte_check(reader, '^');
*flags |= SERD_HAS_DATATYPE;
if ((st = read_iri(reader, datatype, ate_dot))) {
- *datatype = pop_node(reader, *datatype);
- *lang = pop_node(reader, *lang);
- *dest = pop_node(reader, *dest);
return r_err(reader, st, "bad literal\n");
}
break;
@@ -904,6 +891,7 @@ read_literal(SerdReader* reader,
static SerdStatus
read_verb(SerdReader* reader, SerdNode** dest)
{
+ const size_t orig_stack_size = reader->stack.size;
if (peek_byte(reader) == '<') {
return read_IRIREF(reader, dest);
}
@@ -922,13 +910,13 @@ read_verb(SerdReader* reader, SerdNode** dest)
if (!st && node->n_bytes == 1 &&
serd_node_string(node)[0] == 'a' &&
next != ':' && !is_PN_CHARS_BASE((uint32_t)next)) {
- pop_node(reader, *dest);
+ serd_stack_pop_to(&reader->stack, orig_stack_size);
*dest = push_node(reader, SERD_URI, NS_RDF "type", 47);
return SERD_SUCCESS;
} else if (st > SERD_FAILURE ||
read_PrefixedName(reader, *dest, false, &ate_dot) ||
ate_dot) {
- *dest = pop_node(reader, *dest);
+ *dest = NULL;
return r_err(reader, SERD_ERR_BAD_SYNTAX, "bad verb\n");
}
@@ -951,7 +939,6 @@ read_BLANK_NODE_LABEL(SerdReader* reader, SerdNode** dest, bool* ate_dot)
if (is_digit(c) || c == '_') {
push_byte(reader, n, eat_byte_safe(reader, c));
} else if (read_PN_CHARS(reader, n)) {
- *dest = pop_node(reader, n);
return r_err(reader, SERD_ERR_BAD_SYNTAX, "invalid name start\n");
}
@@ -978,7 +965,6 @@ read_BLANK_NODE_LABEL(SerdReader* reader, SerdNode** dest, bool* ate_dot)
reader->seen_genid = true;
} else if (reader->seen_genid &&
buf[reader->bprefix_len] == 'B') {
- *dest = pop_node(reader, n);
return r_err(
reader, SERD_ERR_ID_CLASH,
"found both `b' and `B' blank IDs, prefix required\n");
@@ -1061,9 +1047,7 @@ read_object(SerdReader* reader, ReadContext* ctx, bool emit, bool* ate_dot)
static const char* const XSD_BOOLEAN = NS_XSD "boolean";
static const size_t XSD_BOOLEAN_LEN = 40;
-#ifndef NDEBUG
const size_t orig_stack_size = reader->stack.size;
-#endif
SerdStatus ret = SERD_FAILURE;
@@ -1129,7 +1113,6 @@ read_object(SerdReader* reader, ReadContext* ctx, bool emit, bool* ate_dot)
} else {
if ((ret = read_PrefixedName(reader, o, false, ate_dot))) {
ret = ret > SERD_FAILURE ? ret : SERD_ERR_BAD_SYNTAX;
- pop_node(reader, o);
return r_err(reader, ret, "expected prefixed name\n");
}
}
@@ -1148,9 +1131,7 @@ read_object(SerdReader* reader, ReadContext* ctx, bool emit, bool* ate_dot)
return SERD_SUCCESS;
}
- pop_node(reader, lang);
- pop_node(reader, datatype);
- pop_node(reader, o);
+ serd_stack_pop_to(&reader->stack, orig_stack_size);
#ifndef NDEBUG
assert(reader->stack.size == orig_stack_size);
#endif
@@ -1176,12 +1157,14 @@ read_objectList(SerdReader* reader, ReadContext ctx, bool* ate_dot)
static SerdStatus
read_predicateObjectList(SerdReader* reader, ReadContext ctx, bool* ate_dot)
{
+ const size_t orig_stack_size = reader->stack.size;
+
SerdStatus st = SERD_SUCCESS;
while (!(st = read_verb(reader, &ctx.predicate)) &&
read_ws_star(reader) &&
!(st = read_objectList(reader, ctx, ate_dot))) {
- ctx.predicate = pop_node(reader, ctx.predicate);
if (*ate_dot) {
+ serd_stack_pop_to(&reader->stack, orig_stack_size);
return SERD_SUCCESS;
}
@@ -1191,9 +1174,11 @@ read_predicateObjectList(SerdReader* reader, ReadContext ctx, bool* ate_dot)
read_ws_star(reader);
switch (c = peek_byte(reader)) {
case EOF:
+ serd_stack_pop_to(&reader->stack, orig_stack_size);
return r_err(reader, SERD_ERR_BAD_SYNTAX,
"unexpected end of file\n");
case '.': case ']': case '}':
+ serd_stack_pop_to(&reader->stack, orig_stack_size);
return SERD_SUCCESS;
case ';':
eat_byte_safe(reader, c);
@@ -1202,23 +1187,19 @@ read_predicateObjectList(SerdReader* reader, ReadContext ctx, bool* ate_dot)
} while (c == ';');
if (!ate_semi) {
+ serd_stack_pop_to(&reader->stack, orig_stack_size);
return r_err(reader, SERD_ERR_BAD_SYNTAX, "missing ';' or '.'\n");
}
}
- ctx.predicate = pop_node(reader, ctx.predicate);
+ serd_stack_pop_to(&reader->stack, orig_stack_size);
+ ctx.predicate = 0;
return st;
}
static SerdStatus
-end_collection(SerdReader* reader,
- ReadContext ctx,
- SerdNode* n1,
- SerdNode* n2,
- SerdStatus st)
+end_collection(SerdReader* reader, ReadContext ctx, SerdStatus st)
{
- pop_node(reader, n2);
- pop_node(reader, n1);
*ctx.flags &= ~(unsigned)SERD_LIST_CONT;
if (!st) {
return (eat_byte_check(reader, ')') == ')') ? SERD_SUCCESS
@@ -1244,13 +1225,12 @@ read_collection(SerdReader* reader, ReadContext ctx, SerdNode** dest)
}
if (end) {
- return end_collection(reader, ctx, 0, 0, st);
+ return end_collection(reader, ctx, st);
}
/* The order of node allocation here is necessarily not in stack order,
so we create two nodes and recycle them throughout. */
SerdNode* n1 = push_node_padded(reader, genid_size(reader), SERD_BLANK, "", 0);
- SerdNode* n2 = 0;
SerdNode* node = n1;
SerdNode* rest = 0;
@@ -1264,14 +1244,14 @@ read_collection(SerdReader* reader, ReadContext ctx, SerdNode** dest)
ctx.predicate = reader->rdf_first;
bool ate_dot = false;
if ((st = read_object(reader, &ctx, true, &ate_dot)) || ate_dot) {
- return end_collection(reader, ctx, n1, n2, st);
+ return end_collection(reader, ctx, st);
}
if (!(end = peek_delim(reader, ')'))) {
/* Give rest a new ID. Done as late as possible to ensure it is
used and > IDs generated by read_object above. */
if (!rest) {
- rest = n2 = blank_id(reader); // First pass, push
+ rest = blank_id(reader); // First pass, push
} else {
set_blank_id(reader, rest, genid_size(reader));
}
@@ -1287,7 +1267,7 @@ read_collection(SerdReader* reader, ReadContext ctx, SerdNode** dest)
node = ctx.subject; // invariant
}
- return end_collection(reader, ctx, n1, n2, st);
+ return end_collection(reader, ctx, st);
}
static SerdStatus
@@ -1310,7 +1290,6 @@ read_subject(SerdReader* reader, ReadContext ctx, SerdNode** dest, int* s_type)
}
if (ate_dot) {
- pop_node(reader, *dest);
return r_err(reader, SERD_ERR_BAD_SYNTAX, "subject ends with `.'\n");
}
@@ -1376,7 +1355,6 @@ read_base(SerdReader* reader, bool sparql, bool token)
if (reader->sink->base) {
TRY(st, reader->sink->base(reader->sink->handle, uri));
}
- pop_node(reader, uri);
read_ws_star(reader);
if (!sparql) {
@@ -1405,7 +1383,6 @@ read_prefixID(SerdReader* reader, bool sparql, bool token)
}
if (eat_byte_check(reader, ':') != ':') {
- pop_node(reader, name);
return SERD_ERR_BAD_SYNTAX;
}
@@ -1416,8 +1393,6 @@ read_prefixID(SerdReader* reader, bool sparql, bool token)
if (reader->sink->prefix) {
st = reader->sink->prefix(reader->sink->handle, name, uri);
}
- pop_node(reader, uri);
- pop_node(reader, name);
if (!sparql) {
read_ws_star(reader);
st = eat_byte_check(reader, '.') ? SERD_SUCCESS : SERD_ERR_BAD_SYNTAX;
@@ -1455,8 +1430,10 @@ read_wrappedGraph(SerdReader* reader, ReadContext* ctx)
read_ws_star(reader);
while (peek_byte(reader) != '}') {
- bool ate_dot = false;
- int s_type = 0;
+ const size_t orig_stack_size = reader->stack.size;
+ bool ate_dot = false;
+ int s_type = 0;
+
ctx->subject = 0;
SerdStatus st = read_subject(reader, *ctx, &ctx->subject, &s_type);
if (st) {
@@ -1465,7 +1442,7 @@ read_wrappedGraph(SerdReader* reader, ReadContext* ctx)
return r_err(reader, SERD_ERR_BAD_SYNTAX,
"missing predicate object list\n");
}
- pop_node(reader, ctx->subject);
+ serd_stack_pop_to(&reader->stack, orig_stack_size);
read_ws_star(reader);
if (peek_byte(reader) == '.') {
eat_byte_safe(reader, '.');
@@ -1538,7 +1515,6 @@ read_n3_statement(SerdReader* reader)
TRY(st, read_labelOrSubject(reader, &ctx.graph));
read_ws_star(reader);
TRY(st, read_wrappedGraph(reader, &ctx));
- pop_node(reader, ctx.graph);
ctx.graph = 0;
read_ws_star(reader);
} else if (read_ws_star(reader) && peek_byte(reader) == '{') {
@@ -1547,9 +1523,8 @@ read_n3_statement(SerdReader* reader)
"invalid graph name\n");
}
ctx.graph = ctx.subject;
- ctx.subject = 0;
+ ctx.subject = NULL;
TRY(st, read_wrappedGraph(reader, &ctx));
- pop_node(reader, ctx.graph);
read_ws_star(reader);
} else if ((st = read_triples(reader, ctx, &ate_dot))) {
if (st == SERD_FAILURE && s_type == '[') {
@@ -1582,13 +1557,16 @@ SerdStatus
read_turtleTrigDoc(SerdReader* reader)
{
while (!reader->source.eof) {
- const SerdStatus st = read_n3_statement(reader);
+ const size_t orig_stack_size = reader->stack.size;
+ const SerdStatus st = read_n3_statement(reader);
if (st > SERD_FAILURE) {
if (reader->strict) {
+ serd_stack_pop_to(&reader->stack, orig_stack_size);
return st;
}
skip_until(reader, '\n');
}
+ serd_stack_pop_to(&reader->stack, orig_stack_size);
}
return SERD_SUCCESS;
}
@@ -1598,6 +1576,8 @@ read_nquadsDoc(SerdReader* reader)
{
SerdStatus st = SERD_SUCCESS;
while (!reader->source.eof) {
+ const size_t orig_stack_size = reader->stack.size;
+
SerdStatementFlags flags = 0;
ReadContext ctx = { 0, 0, 0, 0, 0, 0, &flags };
bool ate_dot = false;
@@ -1641,10 +1621,7 @@ read_nquadsDoc(SerdReader* reader)
TRY(st, emit_statement(reader, ctx, ctx.object));
- pop_node(reader, ctx.graph);
- pop_node(reader, ctx.lang);
- pop_node(reader, ctx.datatype);
- pop_node(reader, ctx.object);
+ serd_stack_pop_to(&reader->stack, orig_stack_size);
}
return SERD_SUCCESS;
}
diff --git a/src/reader.c b/src/reader.c
index bb71be83..c3e2e268 100644
--- a/src/reader.c
+++ b/src/reader.c
@@ -71,6 +71,13 @@ SerdNode*
push_node_padded(SerdReader* reader, size_t maxlen,
SerdType type, const char* str, size_t n_bytes)
{
+ // Push a null byte to ensure the previous node was null terminated
+ char* terminator = (char*)serd_stack_push(&reader->stack, 1);
+ if (!terminator) {
+ return NULL;
+ }
+ *terminator = 0;
+
void* mem = serd_stack_push_aligned(
&reader->stack, sizeof(SerdNode) + maxlen + 1, sizeof(SerdNode));
@@ -89,7 +96,7 @@ push_node_padded(SerdReader* reader, size_t maxlen,
#ifdef SERD_STACK_CHECK
reader->allocs = (Ref*)realloc(
reader->allocs, sizeof(reader->allocs) * (++reader->n_allocs));
- reader->allocs[reader->n_allocs - 1] = (mem - reader->stack.buf);
+ reader->allocs[reader->n_allocs - 1] = node;
#endif
return node;
}
@@ -100,21 +107,6 @@ push_node(SerdReader* reader, SerdType type, const char* str, size_t n_bytes)
return push_node_padded(reader, n_bytes, type, str, n_bytes);
}
-SerdNode*
-pop_node(SerdReader* reader, const SerdNode* node)
-{
- if (node && node != reader->rdf_first && node != reader->rdf_rest
- && node != reader->rdf_nil) {
-#ifdef SERD_STACK_CHECK
- SERD_STACK_ASSERT_TOP(reader, node);
- --reader->n_allocs;
-#endif
- char* const top = reader->stack.buf + reader->stack.size;
- serd_stack_pop_aligned(&reader->stack, (size_t)(top - (char*)node));
- }
- return NULL;
-}
-
SerdStatus
emit_statement(SerdReader* reader, ReadContext ctx, SerdNode* o)
{
@@ -193,9 +185,6 @@ serd_reader_free(SerdReader* reader)
return;
}
- pop_node(reader, reader->rdf_nil);
- pop_node(reader, reader->rdf_rest);
- pop_node(reader, reader->rdf_first);
serd_node_free(reader->default_graph);
#ifdef SERD_STACK_CHECK
diff --git a/src/reader.h b/src/reader.h
index 4925dbda..eae96616 100644
--- a/src/reader.h
+++ b/src/reader.h
@@ -94,8 +94,6 @@ size_t genid_size(SerdReader* reader);
SerdNode* blank_id(SerdReader* reader);
void set_blank_id(SerdReader* reader, SerdNode* node, size_t buf_size);
-SerdNode* pop_node(SerdReader* reader, const SerdNode* node);
-
SerdStatus
emit_statement(SerdReader* reader, ReadContext ctx, SerdNode* o);
@@ -150,7 +148,7 @@ static inline SerdStatus
push_byte(SerdReader* reader, SerdNode* node, const int c)
{
assert(c != EOF);
- SERD_STACK_ASSERT_TOP(reader, ref);
+ SERD_STACK_ASSERT_TOP(reader, node);
char* const s = (char*)serd_stack_push(&reader->stack, 1);
if (!s) {
diff --git a/src/stack.h b/src/stack.h
index 174c6378..cbbe6b18 100644
--- a/src/stack.h
+++ b/src/stack.h
@@ -22,6 +22,7 @@
#include <assert.h>
#include <stddef.h>
#include <stdlib.h>
+#include <string.h>
/** An offset to start the stack at. Note 0 is reserved for NULL. */
#define SERD_STACK_BOTTOM sizeof(void*)
@@ -80,40 +81,28 @@ serd_stack_pop(SerdStack* stack, size_t n_bytes)
stack->size -= n_bytes;
}
+static inline void
+serd_stack_pop_to(SerdStack* stack, size_t n_bytes)
+{
+ assert(stack->size >= n_bytes);
+ stack->size = n_bytes;
+}
+
static inline void*
serd_stack_push_aligned(SerdStack* stack, size_t n_bytes, size_t align)
{
- // Push one byte to ensure space for a pad count
- if (!serd_stack_push(stack, 1)) {
- return NULL;
- }
-
// Push padding if necessary
const size_t pad = align - stack->size % align;
if (pad > 0) {
- if (!serd_stack_push(stack, pad)) {
+ void* padding = serd_stack_push(stack, pad);
+ if (!padding) {
return NULL;
}
+ memset(padding, 0, pad);
}
- // Set top of stack to pad count so we can properly pop later
- stack->buf[stack->size - 1] = (char)pad;
-
// Push requested space at aligned location
return serd_stack_push(stack, n_bytes);
}
-static inline void
-serd_stack_pop_aligned(SerdStack* stack, size_t n_bytes)
-{
- // Pop requested space down to aligned location
- serd_stack_pop(stack, n_bytes);
-
- // Get amount of padding from top of stack
- const uint8_t pad = (uint8_t)stack->buf[stack->size - 1];
-
- // Pop padding and pad count
- serd_stack_pop(stack, pad + 1u);
-}
-
#endif // SERD_STACK_H