Leaking problem

Hi,

Is there any update on this memory leak? We have used DayPilot Scheduler extensively throughout our web application and since rollout we have found this leak to be a major problem. I've found using sIEve that {dayPilot}_scroll, {dayPilot}_main tags and the top resource header DIV are being leaked. One instance of each for every time that the daypilot markup that has come down in our UpdatePanel; however I cannot work out a method of releasing them. The worst bit is that it seems to be leaking the entire page's memory along with these tags. This wouldn't be too much of an issue, however we have clients (with extremely old machines) running the application over cytrix. This results in speed issues for every cytrix user as the "client"s are all eating up the cytrix server's memory at the same time.

NB: We're using IE6 and ASP.Net 2.0.

Our reason for using UpdatePanels instead of using DayPilot's callback functions is that these functions do not cope with changes in slot length &resources etc (basically all the markup). So I am wondering whether you plan to make daypilot capable of rendering the markup completely client-side?

Thanks in advance. P.S. I realise my tone in a bit ranty, but apart from the memory leak, Ithink the control is second to none for scheduling. Keep up the good work.

Phil.

Hi,

I managed to find the leaks inthe end. In the latest updates to IE6 MS says that they have fixed the circular reference (from DOM to javascript & vice versa) causing leaks problem, however I then found that that only counts for elements still attached to the document. When an UpdatePanel refreshes, the contents are removed from the DOM and so the leaks persist. I then found this article (NB: I'm new to this JS leak stuff, so the content of this link may be old hat to most) http://javascript.crockford.com/memory/leak.html. I adapted the brute force cleanup code in this article to output a quasi-xpath stringto a textbox containingthe locations of all handlers it finds under the DayPilot markup.

Here's the modified brute force code... supply d = DayPilot DIV, path = "".

function purge(d, path) 
{ 
var a = d.attributes, i, l, n; 
var tb = document.getElementById('<% =Console.ClientID %>'); 

if (a) 
{ 
l = a.length; 
for (i = 0; i < l; i += 1) 
{ 
n = a[i].name; 

if (typeof d[n] === 'function') 
{ 
if (d[n] != null) 
tb.text += path + '.' + n + '()' + '\n'; 

d[n] = null; 
} 
} 
} 

a = d.childNodes; 
if (a) 
{ 
l = a.length; 
for (i = 0; i < l; i += 1) 
purge(d.childNodes[i], ((d.childNodes[i].id != null && d.childNodes[i].id != '') ? d.childNodes[i].tagName + '[@id=\'' + d.childNodes[i].id + '\']' : path + '.' + d.childNodes[i].tagName)); 
} 
}

From the output I then wrote the following cleanupDayPilotHandlers function which should be calledbefore refreshing an update panel containing DayPilot. I guess it would have to be modified if it's to be called before a call back (by the way this leaks too, but only when the number of columns change).

function getFirstChildByTagName(parent, childTagName) 
{ 
var child = parent.firstChild; 
while(child!=null) 
{ 
if (child.tagName == childTagName) 
return child; 

child = child.nextSibling; 
} 

return null; 
} 

//dayPilot is the javaScript dayPilot instance eg. dpc1.
function cleanupHandlers(dayPilot) 
{ 
dayPilot.selectedEvents = []; 
var dayPilotName = dayPilot.id; 

var header = document.getElementById(dayPilotName + '_header'); 
var scroller = document.getElementById(dayPilotName + '_scroll'); 
var main = document.getElementById(dayPilotName + '_main'); 
var events = document.getElementById(dayPilotName + '_events'); 

var headerTBody = getFirstChildByTagName(header, 'TBODY'); 
var headerTR = getFirstChildByTagName(headerTBody, 'TR'); 
headerTD = headerTR.firstChild 
while (headerTD != null) 
{ 
if (headerTD.tagName == 'TD') 
headerTD.onclick = null; 

headerTD = headerTD.nextSibling; 
} 

scroller.onscroll = null; 

main.onmouseup = null; 
main.onmousemove = null; 

eventsTD = events.firstChild; 
while (eventsTD != null) 
{ 
if (eventsTD.tagName == 'TD') 
{ 
var outerDIV = getFirstChildByTagName(eventsTD, 'DIV'); 

var innerDIV = outerDIV.firstChild; 

while (innerDIV != null) 
{ 
if (innerDIV.tagName == 'DIV') 
{ 
innerDIV.data = null; 
innerDIV.event = null; 
innerDIV.oncontextmenu = null; 
innerDIV.onmousemove = null; 
innerDIV.onmousedown = null; 
} 

innerDIV = innerDIV.nextSibling; 
} 
} 

eventsTD = eventsTD.nextSibling; 
} 

var mainTBody = getFirstChildByTagName(main, 'TBODY'); 
var mainTR = mainTBody.firstChild; 
while (mainTR != null) 
{ 
if (mainTR.tagName == 'TR') 
{ 
var mainTD = mainTR.firstChild; 

while(mainTD != null) 
{ 
if (mainTD.tagName == 'TD') 
{ 
mainTD.onmouseup = null; 
mainTD.oncontextmenu = null; 
mainTD.onmousemove = null; 
mainTD.onclick = null; 
mainTD.root = null; 
} 

mainTD = mainTD.nextSibling; 
} 
} 

mainTR = mainTR.nextSibling; 
} 
}

NB: I'm not sure that this is exhaustive as we're only using a limited set of features (basically displaying a schedule, selecting an event and selecting time ranges), so if you find more memory leaks, use the purge function above to track them down.

Regards,

Phil.

Hi Dan,

I am using DaypilotScheduler version 5.0 with Latest Service Pack. I have about about 300 resources and 700 events in a dayview.The IE memory usagestarts with 40 MB and goes upto 1 GB after 5 to 6 hours.The dayview refreshes after every 5 mins.As told by you the leaking problem has been fixed in version 4.9.Any help would be much appriciated.