Only record one in n line endings to save space.

R=yangguo@chromium.org
BUG=

Review URL: https://codereview.chromium.org/1137683003

Cr-Commit-Position: refs/heads/master@{#28837}
This commit is contained in:
erikcorry 2015-06-08 05:00:47 -07:00 committed by Commit bot
parent 5eafd7a3d9
commit b3d4bce593
9 changed files with 360 additions and 122 deletions

View File

@ -2552,8 +2552,7 @@ void Debug::OnCompileError(Handle<Script> script) {
}
void Debug::OnDebugBreak(Handle<Object> break_points_hit,
bool auto_continue) {
void Debug::OnDebugBreak(Handle<Object> break_points_hit, bool auto_continue) {
// The caller provided for DebugScope.
AssertDebugContext();
// Bail out if there is no listener for this event

View File

@ -511,18 +511,9 @@ class CaptureStackTraceHelper {
// line_number is already shifted by the script_line_offset.
int relative_line_number = line_number - script_line_offset;
if (!column_key_.is_null() && relative_line_number >= 0) {
Handle<FixedArray> line_ends(FixedArray::cast(script->line_ends()));
int start = (relative_line_number == 0) ? 0 :
Smi::cast(line_ends->get(relative_line_number - 1))->value() + 1;
int column_offset = position - start;
if (relative_line_number == 0) {
// For the case where the code is on the same line as the script
// tag.
column_offset += script->column_offset()->value();
}
int column = Script::GetColumnNumber(script, position);
JSObject::AddProperty(stack_frame, column_key_,
handle(Smi::FromInt(column_offset + 1), isolate_),
NONE);
handle(Smi::FromInt(column + 1), isolate_), NONE);
}
JSObject::AddProperty(stack_frame, line_key_,
handle(Smi::FromInt(line_number + 1), isolate_),

View File

@ -305,6 +305,13 @@ macro ORDERED_HASH_MAP_CHAIN_AT(table, entry, numBuckets) = (FIXED_ARRAY_GET(tab
# Must match OrderedHashTable::kNotFound.
define NOT_FOUND = -1;
# Line ends array constants - see v8::internal::Script
define REDUCTION_INDEX = 0;
define NUMBER_OF_LINES_INDEX = 1;
define FIRST_LINE_END_INDEX = 2;
define ASCII_NL = 10;
define ASCII_CR = 13;
# Check whether debug is active.
define DEBUG_IS_ACTIVE = (%_DebugIsActive() != 0);
macro DEBUG_IS_STEPPING(function) = (%_DebugIsActive() != 0 && %DebugCallbackSupportsStepping(function));

View File

@ -40,6 +40,7 @@ var InternalArray = utils.InternalArray;
var ObjectDefineProperty = utils.ObjectDefineProperty;
var ArrayJoin;
var MathFloor;
var ObjectToString;
var StringCharAt;
var StringIndexOf;
@ -47,6 +48,7 @@ var StringSubstring;
utils.Import(function(from) {
ArrayJoin = from.ArrayJoin;
MathFloor = from.MathFloor;
ObjectToString = from.ObjectToString;
StringCharAt = from.StringCharAt;
StringIndexOf = from.StringIndexOf;
@ -203,41 +205,94 @@ function GetSourceLine(message) {
return location.sourceText();
}
function Newlines(source, from, to, reduction) {
var newLines = new InternalArray();
if (!IS_STRING(source)) return newLines;
var length = source.length;
for (; from < to && from < length && newLines.length < reduction - 1
; ++from) {
var c = %_StringCharCodeAt(source, from);
if (c == ASCII_CR) {
if (from < length - 1) {
var c2 = %_StringCharCodeAt(source, from + 1);
if (c2 == ASCII_NL) {
from++; // CR-LF counts as one newline.
}
}
newLines.push(from);
} else if (c == ASCII_NL) {
newLines.push(from);
}
}
// End-of-file virtual end-of-line.
if (to >= length) {
var last = length != 0 ? %_StringCharCodeAt(source, length - 1) : 0;
if (last != ASCII_NL && last != ASCII_CR) newLines.push(source.length - 1);
}
return newLines;
}
function ScriptLineEnd(line) {
if (line < 0) return -1;
var source = this.source;
if (!IS_STRING(source)) return -1;
var line_ends = this.line_ends;
var reduction = line_ends[REDUCTION_INDEX];
var index = MathFloor(line / reduction) + FIRST_LINE_END_INDEX;
if (index >= line_ends.length) return -1;
var position = line_ends[index];
if (line % reduction == 0) return position;
var lines = Newlines(source, position + 1, source.length, reduction);
return lines[line % reduction - 1];
}
/**
* Find a line number given a specific source position.
* @param {number} position The source position.
* @return {number} 0 if input too small, -1 if input too large,
else the line number.
* @return {number} -1 if position too large, else the 0-based line number.
*/
function ScriptLineFromPosition(position) {
var lower = 0;
var upper = this.lineCount() - 1;
var source = this.source;
if (!IS_STRING(source)) return -1;
var line_ends = this.line_ends;
var lower = FIRST_LINE_END_INDEX;
var upper = line_ends.length - 1;
// We'll never find invalid positions so bail right away.
if (position > line_ends[upper]) {
return -1;
}
var reduction = line_ends[REDUCTION_INDEX];
// This '>' would normally be a '>=', but due to {}-less 'with' statements in
// top-level code we sometimes encounter code positions that are one character
// after the end of the source. See comment in Rewriter::Rewrite.
if (position > source.length) return -1;
// This means we don't have to safe-guard indexing line_ends[i - 1].
if (position <= line_ends[0]) {
return 0;
}
var index = 0;
// Binary search to find line # from position range.
while (upper >= 1) {
var i = (lower + upper) >> 1;
if (position > line_ends[i]) {
lower = i + 1;
} else if (position <= line_ends[i - 1]) {
upper = i - 1;
if (position > line_ends[upper]) {
index = upper;
} else {
return i;
// Invariant: position > line_ends[lower]
// Invariant: position <= line_ends[upper]
while (lower + 1 < upper) {
// Since they differ by at least 2, i must be different from both
// upper or lower.
var i = (lower + upper) >> 1;
if (position > line_ends[i]) {
lower = i;
} else {
upper = i;
}
}
index = lower;
}
return -1;
var line = (index - FIRST_LINE_END_INDEX) * reduction;
return line +
Newlines(source, line_ends[index] + 1, position, reduction).length;
}
/**
@ -250,14 +305,19 @@ function ScriptLineFromPosition(position) {
*/
function ScriptLocationFromPosition(position,
include_resource_offset) {
// Get zero-based line number.
var line = this.lineFromPosition(position);
if (line == -1) return null;
// Determine start, end and column.
var line_ends = this.line_ends;
var start = line == 0 ? 0 : line_ends[line - 1] + 1;
var end = line_ends[line];
if (end > 0 && %_CallFunction(this.source, end - 1, StringCharAt) == '\r') {
var start = this.lineEnd(line) + 1;
// End will be used for substr, so make it non-inclusive.
var end = this.lineEnd(line + 1) + 1;
if (end > this.source.length) end = this.source.length;
// But trim the newline if there is one (there might not be at EOF).
while (end > start) {
var trim_char = %_CallFunction(this.source, end - 1, StringCharAt);
if (trim_char != '\n' && trim_char != '\r') break;
end--;
}
var column = position - start;
@ -317,7 +377,7 @@ function ScriptLocationFromLine(opt_line, opt_column, opt_offset_position) {
}
return this.locationFromPosition(
this.line_ends[offset_line + line - 1] + 1 + column); // line > 0 here.
this.lineEnd(offset_line + line) + 1 + column); // line > 0 here.
}
}
@ -351,9 +411,8 @@ function ScriptSourceSlice(opt_from_line, opt_to_line) {
return null;
}
var line_ends = this.line_ends;
var from_position = from_line == 0 ? 0 : line_ends[from_line - 1] + 1;
var to_position = to_line == 0 ? 0 : line_ends[to_line - 1] + 1;
var from_position = this.lineEnd(from_line) + 1;
var to_position = this.lineEnd(to_line) + 1;
// Return a source slice with line numbers re-adjusted to the resource.
return new SourceSlice(this,
@ -377,9 +436,8 @@ function ScriptSourceLine(opt_line) {
}
// Return the source line.
var line_ends = this.line_ends;
var start = line == 0 ? 0 : line_ends[line - 1] + 1;
var end = line_ends[line];
var start = this.lineEnd(line) + 1;
var end = this.lineEnd(line + 1);
return %_CallFunction(this.source, start, end, StringSubstring);
}
@ -391,7 +449,7 @@ function ScriptSourceLine(opt_line) {
*/
function ScriptLineCount() {
// Return number of source lines.
return this.line_ends.length;
return this.line_ends[NUMBER_OF_LINES_INDEX];
}
@ -426,7 +484,8 @@ utils.SetUpLockedPrototype(Script, [
"sourceSlice", ScriptSourceSlice,
"sourceLine", ScriptSourceLine,
"lineCount", ScriptLineCount,
"nameOrSourceURL", ScriptNameOrSourceURL
"nameOrSourceURL", ScriptNameOrSourceURL,
"lineEnd", ScriptLineEnd
]
);

View File

@ -8901,18 +8901,62 @@ static void CalculateLineEndsImpl(Isolate* isolate,
Vector<const SourceChar> src,
bool include_ending_line) {
const int src_len = src.length();
bool exotic_newlines = false;
if (include_ending_line) {
// Initally assume reduction is 1, ie all line endings are in the array.
DCHECK_EQ(line_ends->length(), Script::kReductionIndex);
line_ends->Add(1);
// Write a placeholder for the number-of-lines indicator.
DCHECK_EQ(line_ends->length(), Script::kNumberOfLinesIndex);
line_ends->Add(0);
DCHECK_EQ(line_ends->length(), Script::kFirstLineEndIndex);
// There's a fictional newline just before the first character. This
// simplifies a lot of things.
line_ends->Add(-1);
}
UnicodeCache* cache = isolate->unicode_cache();
for (int i = 0; i < src_len - 1; i++) {
SourceChar current = src[i];
SourceChar next = src[i + 1];
if (cache->IsLineTerminatorSequence(current, next)) line_ends->Add(i);
if (cache->IsLineTerminatorSequence(current, next)) {
if (current != '\n' && current != '\r') exotic_newlines = true;
line_ends->Add(i);
}
}
if (src_len > 0 && cache->IsLineTerminatorSequence(src[src_len - 1], 0)) {
line_ends->Add(src_len - 1);
int last_posn = src_len - 1;
if (last_posn >= 0 && cache->IsLineTerminatorSequence(src[last_posn], 0)) {
if (src[last_posn] != '\n' && src[last_posn] != '\r')
exotic_newlines = true;
line_ends->Add(last_posn);
} else if (include_ending_line) {
// Even if the last line misses a line end, it is counted.
line_ends->Add(src_len);
// Even if the last line misses a line end, it is counted. Because we
// sometimes use character positions that are one beyond the end of the
// source (see Rewriter::Rewrite) we set the newline one beyond that.
// This is used for substr calculations, which trims to string length,
// so it's harmless.
line_ends->Add(last_posn + 1);
}
if (include_ending_line) {
// Update number of lines in script.
int lines = line_ends->length() - (Script::kFirstLineEndIndex + 1);
line_ends->Set(Script::kNumberOfLinesIndex, lines);
// Abuse some flags. The bots will run with a good variety of these flags,
// giving better coverage for the reduction code.
bool always_reduce = FLAG_always_opt;
bool never_reduce = !FLAG_crankshaft;
if (!never_reduce && !exotic_newlines &&
(always_reduce ||
(line_ends->length() > 5 && line_ends->length() * 8 > src_len / 12))) {
// If the line-ends array (8 bytes per entry) is larger than about 8%
// of the source length, then we reduce it to save memory. This won't
// trigger if lines are > 100 characters on average. If it triggers, then
// the goal is for it to take only 3% of the source size.
int reduction =
always_reduce ? 2 : (line_ends->length() * 8 * 33 / src_len);
DCHECK(reduction > 1);
line_ends->Set(Script::kReductionIndex, reduction);
}
}
}
@ -8941,13 +8985,31 @@ Handle<FixedArray> String::CalculateLineEnds(Handle<String> src,
include_ending_line);
}
}
int line_count = line_ends.length();
Handle<FixedArray> array = isolate->factory()->NewFixedArray(line_count);
for (int i = 0; i < line_count; i++) {
if (include_ending_line) {
const int kReductionIndex = Script::kReductionIndex;
const int kFirstLineEndIndex = Script::kFirstLineEndIndex;
int line_count = line_ends.length() - kFirstLineEndIndex;
int reduction = line_ends[kReductionIndex];
int reduced_lines = (line_count + reduction - 1) / reduction;
Handle<FixedArray> array =
isolate->factory()->NewFixedArray(kFirstLineEndIndex + reduced_lines);
for (int i = 0; i < kFirstLineEndIndex; i++) {
array->set(i, Smi::FromInt(line_ends[i]));
}
int j = kFirstLineEndIndex;
for (int i = 0; i < line_count; i += reduction, ++j) {
array->set(j, Smi::FromInt(line_ends[i + kFirstLineEndIndex]));
}
return array;
} else {
Handle<FixedArray> array =
isolate->factory()->NewFixedArray(line_ends.length());
for (int i = 0; i < line_ends.length(); i++) {
array->set(i, Smi::FromInt(line_ends[i]));
}
return array;
}
}
// Compares the contents of two strings by reading and comparing
@ -10304,41 +10366,113 @@ void Script::InitLineEnds(Handle<Script> script) {
}
static int CountForwardNNewlines(Handle<Script> script, int block_position,
int n) {
int position = block_position;
Handle<Object> source_object(script->source(), script->GetIsolate());
if (!source_object->IsString() || n == 0) return position;
Handle<String> source(Handle<String>::cast(source_object));
int length = source->length();
for (int i = position; i < length; i++) {
uc16 current = source->Get(i);
if (current == '\r') {
n--;
if (i + 1 < length && source->Get(i + 1) == '\n') i++;
} else if (current == '\n') {
n--;
}
if (n == 0) return i + 1;
}
if (n == 1 && length > 0) {
uc16 last = source->Get(length - 1);
if (last != '\n' && last != '\r') return length;
}
return -1;
}
int Script::GetColumnNumber(Handle<Script> script, int code_pos) {
// Get zero-based line number.
int line_number = GetLineNumber(script, code_pos);
if (line_number == -1) return -1;
DisallowHeapAllocation no_allocation;
FixedArray* line_ends_array = FixedArray::cast(script->line_ends());
line_number = line_number - script->line_offset()->value();
if (line_number == 0) return code_pos + script->column_offset()->value();
int prev_line_end_pos =
Smi::cast(line_ends_array->get(line_number - 1))->value();
return code_pos - (prev_line_end_pos + 1);
int reduction = Smi::cast(line_ends_array->get(kReductionIndex))->value();
int line_block_position =
Smi::cast(line_ends_array->get(line_number / reduction +
kFirstLineEndIndex))->value() +
1;
int line_position = CountForwardNNewlines(script, line_block_position,
line_number % reduction);
if (line_number == 0) line_position = -script->column_offset()->value();
return code_pos - line_position;
}
// Zero-based line number, calculated from UTF16 character position.
int Script::GetLineNumberWithArray(int code_pos) {
DisallowHeapAllocation no_allocation;
DCHECK(line_ends()->IsFixedArray());
FixedArray* line_ends_array = FixedArray::cast(line_ends());
int line_ends_len = line_ends_array->length();
if (line_ends_len == 0) return -1;
if (line_ends_len == 0) return -1; // This happens if there is no source.
// There's always at least one line ending: A fictional newline just before
// the start.
DCHECK_GE(line_ends_len, kFirstLineEndIndex + 1);
int lower = kFirstLineEndIndex;
int upper = line_ends_len - 1;
if ((Smi::cast(line_ends_array->get(0)))->value() >= code_pos) {
return line_offset()->value();
}
if (code_pos < 0) return -1;
int index = 0;
int left = 0;
int right = line_ends_len;
while (int half = (right - left) / 2) {
if ((Smi::cast(line_ends_array->get(left + half)))->value() > code_pos) {
right -= half;
if (code_pos > Smi::cast(line_ends_array->get(upper))->value()) {
index = upper;
} else {
left += half;
while (lower + 1 < upper) {
DCHECK_LE(Smi::cast(line_ends_array->get(lower))->value(), code_pos);
DCHECK_LE(code_pos, Smi::cast(line_ends_array->get(upper))->value());
int i = (lower + upper) >> 1;
DCHECK(lower != i && upper != i);
if ((Smi::cast(line_ends_array->get(i)))->value() >= code_pos) {
upper = i;
} else {
lower = i;
}
}
return right + line_offset()->value();
index = lower;
}
int reduction = Smi::cast(line_ends_array->get(kReductionIndex))->value();
int line_number = (index - kFirstLineEndIndex) * reduction;
// We only saved an nth of the line ends in the array, because there were so
// many.
int start_of_earlier_line =
Smi::cast(line_ends_array->get(index))->value() + 1;
if (reduction == 1 || !source()->IsString()) {
return line_number + line_offset()->value();
}
String* src = String::cast(source());
// This '>' would normally be a '>=', but due to {}-less 'with' statements in
// top-level code we sometimes encounter code positions that are one character
// after the end of the source. See comment in Rewriter::Rewrite.
if (code_pos > src->length()) return -1;
for (int i = start_of_earlier_line; i < src->length() && i < code_pos; i++) {
uc16 current = src->Get(i);
if (current == '\r') {
if (i < code_pos - 1 && i < src->length() - 1 && src->Get(i + 1) == '\n')
i++;
line_number++;
} else if (current == '\n') {
line_number++;
}
}
return line_number + line_offset()->value();
}

View File

@ -6540,6 +6540,11 @@ class Script: public Struct {
static const int kSourceMappingUrlOffset = kSourceUrlOffset + kPointerSize;
static const int kSize = kSourceMappingUrlOffset + kPointerSize;
// Sync with constants in macros.py.
static const int kReductionIndex = 0;
static const int kNumberOfLinesIndex = 1;
static const int kFirstLineEndIndex = 2;
private:
int GetLineNumberWithArray(int code_pos);

View File

@ -235,6 +235,9 @@ bool Rewriter::Rewrite(ParseInfo* info) {
// eval('with ({x:1}) x = 1');
// the end position of the function generated for executing the eval code
// coincides with the end of the with scope which is the position of '1'.
// Note that this may mean the position is outside the source code
// completely if there is no terminal newline, curly brace, or semicolon,
// often the case for 'eval'.
int pos = function->end_position();
VariableProxy* result_proxy =
processor.factory()->NewVariableProxy(result, pos);

View File

@ -14301,6 +14301,32 @@ void AnalyzeStackInNativeCode(const v8::FunctionCallbackInfo<v8::Value>& args) {
}
void ChangeNewlines(int kind, char* dest, size_t dest_len, const char* source) {
if (kind == 0) {
for (size_t i = 0; i <= strlen(source); i++) {
dest[i] = source[i];
}
} else {
for (size_t i = 0; i <= strlen(source); i++) {
char c = source[i];
if (c == '\n') {
if (kind == 1) {
*dest++ = '\r';
*dest++ = '\n';
} else {
// UTF-8 version of 0x2028 newline.
*dest++ = '\xe2';
*dest++ = '\x80';
*dest++ = '\xa8';
}
} else {
*dest++ = c;
}
}
}
}
// Tests the C++ StackTrace API.
// TODO(3074796): Reenable this as a THREADED_TEST once it passes.
// THREADED_TEST(CaptureStackTrace) {
@ -14314,6 +14340,7 @@ TEST(CaptureStackTrace) {
v8::FunctionTemplate::New(isolate, AnalyzeStackInNativeCode));
LocalContext context(0, templ);
for (int i = 0; i < 3; i++) {
// Test getting OVERVIEW information. Should ignore information that is not
// script name, function name, line number, and column offset.
const char* overview_source =
@ -14325,8 +14352,13 @@ TEST(CaptureStackTrace) {
" bar();\n"
"}\n"
"var x;eval('new foo();');";
size_t munged_length = strlen(overview_source) * 3 + 1;
char* overview_munged_source = new char[munged_length];
ChangeNewlines(i, overview_munged_source, munged_length, overview_source);
v8::Handle<v8::String> overview_src =
v8::String::NewFromUtf8(isolate, overview_source);
v8::String::NewFromUtf8(isolate, overview_munged_source);
delete[] overview_munged_source;
v8::ScriptCompiler::Source script_source(overview_src,
v8::ScriptOrigin(origin));
v8::Handle<Value> overview_result(
@ -14345,8 +14377,12 @@ TEST(CaptureStackTrace) {
" bat();\n"
"}\n"
"eval('new baz();');";
munged_length = strlen(detailed_source) * 3 + 1;
char* detailed_munged_source = new char[munged_length];
ChangeNewlines(i, detailed_munged_source, munged_length, detailed_source);
v8::Handle<v8::String> detailed_src =
v8::String::NewFromUtf8(isolate, detailed_source);
v8::String::NewFromUtf8(isolate, detailed_munged_source);
delete[] detailed_munged_source;
// Make the script using a non-zero line and column offset.
v8::Handle<v8::Integer> line_offset = v8::Integer::New(isolate, 3);
v8::Handle<v8::Integer> column_offset = v8::Integer::New(isolate, 5);
@ -14359,6 +14395,7 @@ TEST(CaptureStackTrace) {
CHECK(!detailed_result.IsEmpty());
CHECK(detailed_result->IsObject());
}
}
static void StackTraceForUncaughtExceptionListener(

View File

@ -63,11 +63,11 @@ var comment_lines = 28;
// This is the last position in the entire file (note: this equals
// file size of <debug-sourceinfo.js> - 1, since starting at 0).
var last_position = 11337;
var last_position = 11591;
// This is the last line of entire file (note: starting at 0).
var last_line = 265;
// This is the last column of last line (note: starting at 0 and +1, due
// to trailing <LF>).
var last_line = 268;
// This is the column of the last character (note: starting at 0) due to
// final line having a trailing newline that is conceptually part of that line.
var last_column = 1;
// This magic number is the length or the first line comment (actually number
@ -250,7 +250,10 @@ assertEquals(158 + start_d, Debug.findFunctionSourceLocation(d, 17, 0).position)
// Make sure invalid inputs work properly.
assertEquals(0, script.locationFromPosition(-1).line);
assertEquals(null, script.locationFromPosition(last_position + 1));
// We might expect last_position + 1 to be the first illegal position, but we
// sometimes generate character positions that are one past the last character.
// See Rewriter::Rewrite for details.
assertEquals(null, script.locationFromPosition(last_position + 2));
// Test last position.
assertEquals(last_position, script.locationFromPosition(last_position).position);