Performance Enhancements

- Nibbler: Now locally stores input length rather than repeatedly calling
  std::string::length.
- Nibbler: Makes greater use of std::string::find_first_not_of, instead of
  looping.
- TDB: No longer applies empty filters to lists - just copies the lists.
- TDB: Now caches data when reading completed.data.
- TDB: During loadPending and loadCompleted, makes fewer copies of the data.
- TDB: In commit, breaks out of search loops after finding the right data.
- TDB: In gc, only writes out minimal pending or completed data, instead of
       all data, all the time.
- TDB: No longer reads completed.data in gc, and simply appends completed
       and deleted tasks to it.
This commit is contained in:
Paul Beckingham 2010-06-23 17:11:59 -04:00
parent cd648270ab
commit abf31a6b35
6 changed files with 179 additions and 104 deletions

1
NEWS
View file

@ -8,6 +8,7 @@ New Features in task 1.9
- New 'show' command to display configuration settings.
- New 'denotate' command to delete annotations.
- New limit:page filter to show only one page of tasks.
- Performance enhancements.
Please refer to the ChangeLog file for full details. There are too many to
list here.

View file

@ -26,12 +26,16 @@
////////////////////////////////////////////////////////////////////////////////
#include <stdlib.h>
#include <string.h>
#include <ctype.h>
#include "Nibbler.h"
const char* c_digits = "0123456789";
////////////////////////////////////////////////////////////////////////////////
Nibbler::Nibbler ()
: mInput ("")
, mLength (0)
, mCursor (0)
{
}
@ -39,6 +43,7 @@ Nibbler::Nibbler ()
////////////////////////////////////////////////////////////////////////////////
Nibbler::Nibbler (const char* input)
: mInput (input)
, mLength (strlen (input))
, mCursor (0)
{
}
@ -46,6 +51,7 @@ Nibbler::Nibbler (const char* input)
////////////////////////////////////////////////////////////////////////////////
Nibbler::Nibbler (const std::string& input)
: mInput (input)
, mLength (input.length ())
, mCursor (0)
{
}
@ -54,6 +60,7 @@ Nibbler::Nibbler (const std::string& input)
Nibbler::Nibbler (const Nibbler& other)
{
mInput = other.mInput;
mLength = other.mLength;
mCursor = other.mCursor;
}
@ -63,6 +70,7 @@ Nibbler& Nibbler::operator= (const Nibbler& other)
if (this != &other)
{
mInput = other.mInput;
mLength = other.mLength;
mCursor = other.mCursor;
}
@ -78,7 +86,7 @@ Nibbler::~Nibbler ()
// Extract up until the next c, or EOS.
bool Nibbler::getUntil (char c, std::string& result)
{
if (mCursor < mInput.length ())
if (mCursor < mLength)
{
std::string::size_type i = mInput.find (c, mCursor);
if (i != std::string::npos)
@ -89,7 +97,7 @@ bool Nibbler::getUntil (char c, std::string& result)
else
{
result = mInput.substr (mCursor);
mCursor = mInput.length ();
mCursor = mLength;
}
return true;
@ -101,7 +109,7 @@ bool Nibbler::getUntil (char c, std::string& result)
////////////////////////////////////////////////////////////////////////////////
bool Nibbler::getUntil (const std::string& terminator, std::string& result)
{
if (mCursor < mInput.length ())
if (mCursor < mLength)
{
std::string::size_type i = mInput.find (terminator, mCursor);
if (i != std::string::npos)
@ -112,7 +120,7 @@ bool Nibbler::getUntil (const std::string& terminator, std::string& result)
else
{
result = mInput.substr (mCursor);
mCursor = mInput.length ();
mCursor = mLength;
}
return true;
@ -124,7 +132,7 @@ bool Nibbler::getUntil (const std::string& terminator, std::string& result)
////////////////////////////////////////////////////////////////////////////////
bool Nibbler::getUntilOneOf (const std::string& chars, std::string& result)
{
if (mCursor < mInput.length ())
if (mCursor < mLength)
{
std::string::size_type i = mInput.find_first_of (chars, mCursor);
if (i != std::string::npos)
@ -135,7 +143,7 @@ bool Nibbler::getUntilOneOf (const std::string& chars, std::string& result)
else
{
result = mInput.substr (mCursor);
mCursor = mInput.length ();
mCursor = mLength;
}
return true;
@ -147,10 +155,10 @@ bool Nibbler::getUntilOneOf (const std::string& chars, std::string& result)
////////////////////////////////////////////////////////////////////////////////
bool Nibbler::skipN (const int quantity /* = 1 */)
{
if (mCursor >= mInput.length ())
if (mCursor >= mLength)
return false;
if (mCursor <= mInput.length () - quantity)
if (mCursor <= mLength - quantity)
{
mCursor += quantity;
return true;
@ -162,7 +170,7 @@ bool Nibbler::skipN (const int quantity /* = 1 */)
////////////////////////////////////////////////////////////////////////////////
bool Nibbler::skip (char c)
{
if (mCursor < mInput.length () &&
if (mCursor < mLength &&
mInput[mCursor] == c)
{
++mCursor;
@ -175,13 +183,17 @@ bool Nibbler::skip (char c)
////////////////////////////////////////////////////////////////////////////////
bool Nibbler::skipAll (char c)
{
std::string::size_type i = mCursor;
while (i < mInput.length () && mInput[i] == c)
++i;
if (i != mCursor)
if (mCursor < mLength)
{
std::string::size_type i = mInput.find_first_not_of (c, mCursor);
if (i == mCursor)
return false;
if (i == std::string::npos)
mCursor = mLength; // Yes, off the end.
else
mCursor = i;
return true;
}
@ -191,14 +203,14 @@ bool Nibbler::skipAll (char c)
////////////////////////////////////////////////////////////////////////////////
bool Nibbler::skipAllOneOf (const std::string& chars)
{
if (mCursor < mInput.length ())
if (mCursor < mLength)
{
std::string::size_type i = mInput.find_first_not_of (chars, mCursor);
if (i == mCursor)
return false;
if (i == std::string::npos)
mCursor = mInput.length (); // Yes, off the end.
mCursor = mLength; // Yes, off the end.
else
mCursor = i;
@ -212,10 +224,10 @@ bool Nibbler::skipAllOneOf (const std::string& chars)
bool Nibbler::getQuoted (char c, std::string& result)
{
std::string::size_type start = mCursor;
if (start < mInput.length () && mInput[start] == c)
if (start < mLength && mInput[start] == c)
{
++start;
if (start < mInput.length ())
if (start < mLength)
{
std::string::size_type end = mInput.find (c, start);
if (end != std::string::npos)
@ -235,7 +247,7 @@ bool Nibbler::getInt (int& result)
{
std::string::size_type i = mCursor;
if (i < mInput.length ())
if (i < mLength)
{
if (mInput[i] == '-')
++i;
@ -243,7 +255,8 @@ bool Nibbler::getInt (int& result)
++i;
}
while (i < mInput.length () && isdigit (mInput[i]))
// TODO Potential for use of find_first_not_of
while (i < mLength && isdigit (mInput[i]))
++i;
if (i > mCursor)
@ -260,7 +273,8 @@ bool Nibbler::getInt (int& result)
bool Nibbler::getUnsignedInt (int& result)
{
std::string::size_type i = mCursor;
while (i < mInput.length () && isdigit (mInput[i]))
// TODO Potential for use of find_first_not_of
while (i < mLength && isdigit (mInput[i]))
++i;
if (i > mCursor)
@ -282,10 +296,10 @@ bool Nibbler::getUntilEOL (std::string& result)
////////////////////////////////////////////////////////////////////////////////
bool Nibbler::getUntilEOS (std::string& result)
{
if (mCursor < mInput.length ())
if (mCursor < mLength)
{
result = mInput.substr (mCursor);
mCursor = mInput.length ();
mCursor = mLength;
return true;
}
@ -295,7 +309,7 @@ bool Nibbler::getUntilEOS (std::string& result)
////////////////////////////////////////////////////////////////////////////////
bool Nibbler::depleted ()
{
if (mCursor >= mInput.length ())
if (mCursor >= mLength)
return true;
return false;

View file

@ -55,6 +55,7 @@ public:
private:
std::string mInput;
std::string::size_type mLength;
std::string::size_type mCursor;
};

View file

@ -219,6 +219,7 @@ int TDB::loadPending (std::vector <Task>& tasks, Filter& filter)
try
{
// Only load if not already loaded.
if (mPending.size () == 0)
{
mId = 1;
@ -247,13 +248,23 @@ int TDB::loadPending (std::vector <Task>& tasks, Filter& filter)
}
// Now filter and return.
if (filter.size ())
{
foreach (task, mPending)
if (filter.pass (*task))
tasks.push_back (*task);
}
else
{
foreach (task, mPending)
tasks.push_back (*task);
}
// Hand back any accumulated additions, if TDB::loadPending is being called
// repeatedly.
int fakeId = mId;
if (filter.size ())
{
foreach (task, mNew)
{
task->id = fakeId++;
@ -261,6 +272,15 @@ int TDB::loadPending (std::vector <Task>& tasks, Filter& filter)
tasks.push_back (*task);
}
}
else
{
foreach (task, mNew)
{
task->id = fakeId++;
tasks.push_back (*task);
}
}
}
catch (std::string& e)
{
@ -281,9 +301,11 @@ int TDB::loadCompleted (std::vector <Task>& tasks, Filter& filter)
Timer t ("TDB::loadCompleted");
std::string file;
int line_number;
int line_number = 1;
try
{
if (mCompleted.size () == 0)
{
char line[T_LINE_MAX];
foreach (location, mLocations)
@ -299,14 +321,10 @@ int TDB::loadCompleted (std::vector <Task>& tasks, Filter& filter)
{
// TODO Add hidden attribute indicating source?
if (line[length - 1] == '\n')
line[length - 1] = '\0';
Task task (line);
task.id = 0; // Need a value, just not a valid value.
if (filter.pass (task))
tasks.push_back (task);
mCompleted.push_back (task);
}
++line_number;
@ -314,6 +332,20 @@ int TDB::loadCompleted (std::vector <Task>& tasks, Filter& filter)
}
}
// Now filter and return.
if (filter.size ())
{
foreach (task, mCompleted)
if (filter.pass (*task))
tasks.push_back (*task);
}
else
{
foreach (task, mCompleted)
tasks.push_back (*task);
}
}
catch (std::string& e)
{
std::stringstream s;
@ -329,7 +361,6 @@ int TDB::loadCompleted (std::vector <Task>& tasks, Filter& filter)
void TDB::add (const Task& task)
{
mNew.push_back (task);
}
////////////////////////////////////////////////////////////////////////////////
@ -344,7 +375,7 @@ void TDB::update (const Task& task)
int TDB::commit ()
{
Timer t ("TDB::commit");
context.hooks.trigger ("pre-gc");
context.hooks.trigger ("pre-commit");
int quantity = mNew.size () + mModified.size ();
@ -364,7 +395,7 @@ int TDB::commit ()
writeUndo (*task, mLocations[0].undo);
mNew.clear ();
context.hooks.trigger ("post-gc");
context.hooks.trigger ("post-commit");
return quantity;
}
@ -375,10 +406,20 @@ int TDB::commit ()
// new tasks appended.
std::vector <Task> allPending;
allPending = mPending;
foreach (task, allPending)
foreach (mtask, mModified)
{
foreach (task, allPending)
{
if (task->id == mtask->id)
{
*task = *mtask;
goto next_mod;
}
}
next_mod:
;
}
foreach (task, mNew)
allPending.push_back (*task);
@ -396,10 +437,20 @@ int TDB::commit ()
// Update the undo log.
if (fseek (mLocations[0].undo, 0, SEEK_END) == 0)
{
foreach (task, mPending)
foreach (mtask, mModified)
{
foreach (task, mPending)
{
if (task->id == mtask->id)
{
writeUndo (*task, *mtask, mLocations[0].undo);
goto next_mod2;
}
}
next_mod2:
;
}
foreach (task, mNew)
writeUndo (*task, mLocations[0].undo);
@ -411,7 +462,7 @@ int TDB::commit ()
mNew.clear ();
}
context.hooks.trigger ("post-gc");
context.hooks.trigger ("post-commit");
return quantity;
}
@ -423,43 +474,45 @@ int TDB::gc ()
{
Timer t ("TDB::gc");
int count = 0;
int count_pending_changes = 0;
int count_completed_changes = 0;
Date now;
// Set up a second TDB.
if (mNew.size ())
throw std::string ("Unexpected new tasks found during gc.");
if (mModified.size ())
throw std::string ("Unexpected modified tasks found during gc.");
lock ();
Filter filter;
TDB tdb;
tdb.location (mLocations[0].path);
tdb.lock ();
std::vector <Task> pending;
tdb.loadPending (pending, filter);
std::vector <Task> completed;
tdb.loadCompleted (completed, filter);
std::vector <Task> ignore;
loadPending (ignore, filter);
// Now move completed and deleted tasks from the pending list to the
// completed list. Isn't garbage collection easy?
std::vector <Task> still_pending;
foreach (task, pending)
std::vector <Task> newly_completed;
foreach (task, mPending)
{
std::string st = task->get ("status");
Task::status s = task->getStatus ();
if (s == Task::completed ||
s == Task::deleted)
{
completed.push_back (*task);
++count;
newly_completed.push_back (*task);
++count_pending_changes; // removal
++count_completed_changes; // addition
}
else if (s == Task::waiting)
{
// Wake up tasks that are waiting.
// Wake up tasks that need to be woken.
Date wait_date (atoi (task->get ("wait").c_str ()));
if (now > wait_date)
{
task->setStatus (Task::pending);
task->remove ("wait");
++count;
++count_pending_changes; // modification
}
still_pending.push_back (*task);
@ -468,37 +521,38 @@ int TDB::gc ()
still_pending.push_back (*task);
}
pending = still_pending;
// No commit - all updates performed manually.
if (count > 0)
if (count_pending_changes > 0)
{
if (fseek (tdb.mLocations[0].pending, 0, SEEK_SET) == 0)
if (fseek (mLocations[0].pending, 0, SEEK_SET) == 0)
{
if (ftruncate (fileno (tdb.mLocations[0].pending), 0))
if (ftruncate (fileno (mLocations[0].pending), 0))
throw std::string ("Failed to truncate pending.data file ");
foreach (task, pending)
fputs (task->composeF4 ().c_str (), tdb.mLocations[0].pending);
foreach (task, still_pending)
fputs (task->composeF4 ().c_str (), mLocations[0].pending);
// Update cached copy.
mPending = still_pending;
}
}
if (fseek (tdb.mLocations[0].completed, 0, SEEK_SET) == 0)
// Append the new_completed tasks to completed.data. No need to write out the
// whole list.
if (count_completed_changes > 0)
{
if (ftruncate (fileno (tdb.mLocations[0].completed), 0))
throw std::string ("Failed to truncate completed.data file ");
foreach (task, completed)
fputs (task->composeF4 ().c_str (), tdb.mLocations[0].completed);
}
fseek (mLocations[0].completed, 0, SEEK_END);
foreach (task, newly_completed)
fputs (task->composeF4 ().c_str (), mLocations[0].completed);
}
// Close files.
tdb.unlock ();
unlock ();
std::stringstream s;
s << "gc " << count << " tasks";
s << "gc " << (count_pending_changes + count_completed_changes) << " tasks";
context.debug (s.str ());
return count;
return count_pending_changes + count_completed_changes;
}
////////////////////////////////////////////////////////////////////////////////

View file

@ -75,11 +75,9 @@ private:
int mId;
std::vector <Task> mPending; // Contents of pending.data
std::vector <Task> mCompleted; // Contents of pending.data
std::vector <Task> mNew; // Uncommitted new tasks
std::vector <Task> mModified; // Uncommitted modified tasks
// TODO Need cache of raw file contents to preserve comments.
};
#endif

View file

@ -71,17 +71,19 @@ int main (int argc, char** argv)
Task task ("[name:\"value\"]");
tdb.add (task);
tdb.unlock ();
// P0 C0 N1 M0
pending.clear ();
completed.clear ();
get (pending, completed);
t.ok (pending.size () == 0, "TDB add -> no commit -> empty");
t.ok (completed.size () == 0, "TDB add -> no commit -> empty");
// Add with commit.
tdb.lock ();
tdb.add (task);
tdb.commit ();
tdb.add (task); // P0 C0 N1 M0
tdb.commit (); // P1 C0 N0 M0
tdb.unlock ();
get (pending, completed);
@ -91,36 +93,40 @@ int main (int argc, char** argv)
t.ok (completed.size () == 0, "TDB add -> commit -> saved");
// Update with commit.
tdb.lock ();
pending.clear ();
completed.clear ();
tdb.lock ();
tdb.load (all, context.filter);
all[0].set ("name", "value2");
tdb.update (all[0]);
tdb.commit ();
tdb.update (all[0]); // P1 C0 N0 M1
tdb.commit (); // P1 C0 N0 M0
tdb.unlock ();
pending.clear ();
completed.clear ();
get (pending, completed);
t.ok (all.size () == 1, "TDB update -> commit -> saved");
t.is (all[0].get ("name"), "value2", "TDB load name=value2");
t.is (all[0].id, 1, "TDB load verification id=1");
// GC.
tdb.lock ();
all.clear ();
tdb.lock ();
tdb.loadPending (all, context.filter);
all[0].setStatus (Task::completed);
tdb.update (all[0]);
tdb.update (all[0]); // P1 C0 N0 M1
Task t2 ("[foo:\"bar\" status:\"pending\"]");
tdb.add (t2);
tdb.add (t2); // P1 C0 N1 M1
tdb.commit ();
tdb.unlock ();
pending.clear ();
completed.clear ();
get (pending, completed);
t.is (pending.size (), (size_t)2, "TDB before gc pending #2");
t.is (pending[0].id, 1, "TDB before gc pending id 1");
t.is (pending[0].getStatus (), Task::completed, "TDB before gc pending status completed");
@ -128,11 +134,12 @@ int main (int argc, char** argv)
t.is (pending[1].getStatus (), Task::pending, "TDB before gc pending status pending");
t.is (completed.size (), (size_t)0, "TDB before gc completed 0");
tdb.gc ();
tdb.gc (); // P1 C1 N0 M0
pending.clear ();
completed.clear ();
get (pending, completed);
t.is (pending.size (), (size_t)1, "TDB after gc pending #1");
t.is (pending[0].id, 1, "TDB after gc pending id 2");
t.is (pending[0].getStatus (), Task::pending, "TDB after gc pending status pending");