Interesting... I wonder if this is why I've been having problems with KP's PC2 reader with huge files just over 2GB? I ended up splitting my meshes to work around it, but now I wonder...
On Sat, Nov 16, 2013 at 2:18 PM, Alok Gandhi <[email protected]>wrote: > Hi All, > > I have update the ReadPC2ICE node to version 1.1 > > You can find the addon here <http://bit.ly/1axgUea> and code > here<http://bit.ly/17ZKRpU> > . > > Change Log: > > Bug fixes: > > > > · Point data for files over 2 GB was not read correctly due to > limitations of <int> data type. It is changed to <__int64> which can handle > the data > 2 GB. Now the node can easily read files over 2 GB. > > > > *Notes* > > · The same changes needs to be applied for *KP_PointCacheReader*for > this bugfix. I am not aware if Kai has fixed this already in other > updates to his code. But the > one<http://sculptwork.com/rr/bak/kaipirinha/KP_PointCacheV25.zip> I > had from rray.de does not have this change. Kai, if you are listening can > you confirm that? > > > > · The code for this bugfix is not supported on Linux. However, in > case you want to compile for Linux, please note that the <__int64> data > type for windows translates to <long long> on a gcc compiler for Linux. I > will add support for Linux once I have a machine with Linux up and running. > Till then you have to change the code yourself. > > > > > > · Suppressed unnecessary warning messages in cases of pc2 file > not specified and unable to open file. They were kind of annoying. Just > uncomment my code lines to bring them back if you want. > > > > > > New features: > > > > · When working with pc2 files I always missed the ability to know > beforehand the start and end frame of the file. Now the node supports this. > There are two new output ports that furnish the start and end frames as > scalars. You can use this to do time warps, offsets, view them in viewports > as custom attributes or whatever else you might find it useful for. > > Please feel free to mail me if you have any question or problems/bugs with > this addon. > > Thanks. > > > > On Mon, Nov 4, 2013 at 9:00 PM, Alok Gandhi <[email protected]>wrote: > >> And here is the github for the code: >> >> https://github.com/alok1974/KP_PointCache-Reader-ICE-Node >> >> >> On Mon, Nov 4, 2013 at 8:42 PM, Alok Gandhi <[email protected]>wrote: >> >>> Here is the code for the pc2 reader: >>> >>> ---------------------------------------------------------- >>> """ >>> .PC2 File Format: >>> You can create or modify PointCache2 files by hand using the following >>> file format. >>> The start of the file is a header containing: >>> >>> char cacheSignature[12]; // Will be 'POINTCACHE2' followed by a >>> trailing null character. >>> int fileVersion; // Currently 1 >>> int numPoints; // Number of points per sample >>> float startFrame; // Corresponds to the UI value of the same >>> name. >>> float sampleRate; // Corresponds to the UI value of the same >>> name. >>> int numSamples; // Defines how many samples are stored in >>> the file. >>> >>> Be sure to check the version number. If it isn't 1, then don't mess with >>> the file, >>> as the format will change in the future. >>> >>> Following the header, there is a straight dump of all the cache >>> samples (which are snapshots of all the point positions for an object). >>> Each sample is stored one after the other as a flat array of x/y/z floats >>> for each point (so each sample is (numPoints * sizeof(float) * 3) bytes). >>> """ >>> >>> >>> # Note for Alan : You might want to extend this class by adding a method >>> to fetch the frame data >>> # for a prticular frame. It is trivial to do so, just >>> look at my code below for >>> # getting bounding box data. >>> >>> import os >>> import sys >>> import time >>> from struct import unpack >>> >>> class CacheObject(object): >>> def __init__(self, pth): >>> self._dFile = None >>> self.fPath = pth >>> self._headerString = '' >>> self._cacheFileVersionNumber = 0 >>> self._pCount = 0 >>> self._startFrame = 0 >>> self._sampleRate = 0 >>> self._numSamples = 0 >>> self._mshBBox = {} >>> self._headerRead = False >>> self._dataRead = False >>> >>> def _setFile(self): >>> if not self._dFile: >>> self._dFile = open(self.fPath, 'rb') >>> self._msh = os.path.splitext(os.path.basename(self.fPath))[0] >>> >>> if not self._headerRead: >>> self._processHeader() >>> >>> >>> >>> def _processHeader(self): >>> self._headerRead = True >>> >>> self._setFile() >>> >>> dfile = self._dFile >>> >>> self._headerString = unpack('12s', dfile.read(12))[0] >>> self._cacheFileVersionNumber = unpack('I', dfile.read(4))[0] >>> self._pCount = unpack('L', dfile.read(4))[0] >>> self._startFrame = unpack('f', dfile.read(4))[0] >>> self._sampleRate = unpack('f', dfile.read(4))[0] >>> self._numSamples = unpack('L', dfile.read(4))[0] >>> >>> if not self._dataRead: >>> dfile.flush() >>> dfile.close() >>> self._dFile = None >>> >>> >>> >>> def _processBBoxData(self): >>> self._dataRead = True >>> self._setFile() >>> >>> dfile = self._dFile >>> data = {} >>> >>> >>> s = int(self._startFrame) >>> e = int(s + self._numSamples) >>> >>> for f in range(s, e): >>> xArr = [] >>> yArr = [] >>> zArr = [] >>> >>> for i in range(self._pCount): >>> x = unpack('f', dfile.read(4))[0] >>> y = unpack('f', dfile.read(4))[0] >>> z = unpack('f', dfile.read(4))[0] >>> >>> xArr.append((x, i)) >>> yArr.append((y, i)) >>> zArr.append((z, i)) >>> >>> # min\max data >>> d = (max(xArr)[0], max(yArr)[0], max(zArr)[0], min(xArr)[0], >>> min(yArr)[0], min(zArr)[0]) >>> >>> data[f] = { 0:(d[3], d[4], d[5]), >>> 1:(d[3], d[1], d[5]), >>> 2:(d[0], d[1], d[5]), >>> 3:(d[0], d[4], d[5]), >>> 4:(d[3], d[4], d[2]), >>> 5:(d[3], d[1], d[2]), >>> 6:(d[0], d[1], d[2]), >>> 7:(d[0], d[4], d[2]), >>> } >>> >>> self._mshBBox = data >>> >>> dfile.flush() >>> dfile.close() >>> self._dFile = None >>> >>> def getHeaderInfo(self): >>> self._processHeader() >>> return {'HEADER': self._headerString, >>> 'VERSION': self._cacheFileVersionNumber, >>> 'NB_POINTS': self._pCount, >>> 'START_FRAME': self._startFrame, >>> 'SAMPLE_RATE': self._sampleRate, >>> 'NB_SAMPLES': self._numSamples,} >>> >>> def getBBoxData(self): >>> self._processBBoxData() >>> return self._mshBBox >>> >>> if __name__ == '__main__': >>> f = r'<pc2 file path>' >>> o = CacheObject(f) >>> print o.getHeaderInfo() >>> >>> --------------------------------------------------------- >>> >>> >>> On Mon, Nov 4, 2013 at 8:37 PM, Alok Gandhi >>> <[email protected]>wrote: >>> >>>> Hmm, there are a few options to load the data in reader operator. >>>> Stream, Frame or Memory. Try selecting the frame or memory options. By >>>> default, it is the stream which might have problems similar to one you >>>> have. Anyways, in addition to the ICE Node, few years back, I also wrote a >>>> python reader class to read the pc2 file data directly through python. Not >>>> optimised using numpy or scipy but it can still let you investigate the >>>> contents of a ,pc2 file in a human-readable format. I will also post the >>>> code to it. >>>> >>>> >>>> On Mon, Nov 4, 2013 at 8:33 PM, Alan Fregtman >>>> <[email protected]>wrote: >>>> >>>>> Intermittent empty frames. Sometimes they work, sometimes not. Mind >>>>> you... it's very dense topo, and the pc2 file is about 2GB. It might be >>>>> reaching some sort of limit somewhere. >>>>> >>>>> >>>>> >>>>> >>>>> On Mon, Nov 4, 2013 at 8:29 PM, Alok Gandhi <[email protected] >>>>> > wrote: >>>>> >>>>>> You do realise that my node is an exact copy of the KP_PointCache >>>>>> Reader Operator. Only difference is that instead of applying the >>>>>> pointposition on the host mesh posarray, it furnishes the same data in >>>>>> the >>>>>> vector array format in ICE. >>>>>> >>>>>> I am not sure if reading the pointpositions in ICE through my node >>>>>> will give you any extra functionality that Kai's original operator can't. >>>>>> But anyways, feel free to try it. >>>>>> >>>>>> Btw, may I know what is the issue that you are having ? >>>>>> >>>>>> >>>>>> On Mon, Nov 4, 2013 at 8:25 PM, Alan Fregtman < >>>>>> [email protected]> wrote: >>>>>> >>>>>>> Great! Thank you. Just today we've been experiencing an issue with >>>>>>> KP's reader for this one specific mesh. We're temporarily using Alembic >>>>>>> manually, but I'm curious if your reader will be any better. >>>>>>> >>>>>>> >>>>>>> >>>>>>> On Mon, Nov 4, 2013 at 7:45 PM, Alok Gandhi < >>>>>>> [email protected]> wrote: >>>>>>> >>>>>>>> Surething, will post the source probably on github. My old laptop >>>>>>>> with ubuntu died a few months ago :( so I cant compile it myself. >>>>>>>> >>>>>>>> I will post the link here soon. >>>>>>>> >>>>>>>> Sent from my iPhone >>>>>>>> >>>>>>>> On Nov 4, 2013, at 7:20 PM, Alan Fregtman <[email protected]> >>>>>>>> wrote: >>>>>>>> >>>>>>>> Hey Alok, >>>>>>>> >>>>>>>> Any chance for a Linux compile? Or sources to attempt a compile >>>>>>>> ourselves? >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> On Mon, Nov 4, 2013 at 11:21 AM, Alok Gandhi < >>>>>>>> [email protected]> wrote: >>>>>>>> >>>>>>>>> Glad it helped you. And yea, now thinking in retrospect, you won't >>>>>>>>> need the switch context as the getpointid already does that. >>>>>>>>> >>>>>>>>> Sent from my iPhone >>>>>>>>> >>>>>>>>> On Nov 4, 2013, at 9:23 AM, Cristobal Infante <[email protected]> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> VERY HANDY Alok, I can confirm you don't need the switch context >>>>>>>>> node, all the rest worked fine.. >>>>>>>>> >>>>>>>>> By the way, the BIG difference in relation to the default "Cache >>>>>>>>> on File" read node, is that you are able to move your cached geometry >>>>>>>>> around the scene. So if you have a last minute layout change you >>>>>>>>> can deal with it in rendering. This is way we've stuck with KP op >>>>>>>>> reader so far.. >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> On 4 November 2013 13:58, Alok Gandhi >>>>>>>>> <[email protected]>wrote: >>>>>>>>> >>>>>>>>>> Hi Cristobal, >>>>>>>>>> >>>>>>>>>> The node output a single array of pointpositions. So you have to >>>>>>>>>> convert the array to per point attribute. A getpointid plugged into >>>>>>>>>> the >>>>>>>>>> select in array and them set pointposition will do the trick. Of >>>>>>>>>> course you >>>>>>>>>> know that the target mesh or pointcloud should have same number >>>>>>>>>> points as >>>>>>>>>> in the pc2 file. Also you might need a switch context node before >>>>>>>>>> set >>>>>>>>>> pointposition. >>>>>>>>>> >>>>>>>>>> In case of an empty pointcloud, it is easier. Just plug the >>>>>>>>>> output of the node directly into an add points node. >>>>>>>>>> >>>>>>>>>> >>>>>>>>>> Sent from my iPhone >>>>>>>>>> >>>>>>>>>> On Nov 4, 2013, at 8:13 AM, Cristobal Infante <[email protected]> >>>>>>>>>> wrote: >>>>>>>>>> >>>>>>>>>> Hi Alok, >>>>>>>>>> >>>>>>>>>> Thanks again for sharing this tool, we still rely KP pc2 reader >>>>>>>>>> so having an alternative is really handy. >>>>>>>>>> >>>>>>>>>> I was guessing "Read PC2 File" > "Set Point Position"? >>>>>>>>>> >>>>>>>>>> But I am getting a structure mismatch, probably doing the wrong >>>>>>>>>> thing! >>>>>>>>>> >>>>>>>>>> Thanks, >>>>>>>>>> Cris >>>>>>>>>> >>>>>>>>>> >>>>>>>>>> On 4 November 2013 05:40, Alok Gandhi >>>>>>>>>> <[email protected]>wrote: >>>>>>>>>> >>>>>>>>>>> Ah I see, I was confused when you said " i haven't tried >>>>>>>>>>> multiple files per frame (does softimage even export pc2 files this >>>>>>>>>>> way?) >>>>>>>>>>> ". >>>>>>>>>>> >>>>>>>>>>> Well in that, sure you can export single file per frame per >>>>>>>>>>> object through KP_PointCache manager, you simply have to select the >>>>>>>>>>> start >>>>>>>>>>> and end frame as the same. PC2 file format have the notion of >>>>>>>>>>> "samples" >>>>>>>>>>> rather than frame, so if you have one sample per frame set then >>>>>>>>>>> basically >>>>>>>>>>> you are exporting one data set (pointpositions) per frame. >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> On Mon, Nov 4, 2013 at 12:26 AM, Steven Caron >>>>>>>>>>> <[email protected]>wrote: >>>>>>>>>>> >>>>>>>>>>>> no one said that you could have multiple objects in the same >>>>>>>>>>>> .pc2 file. >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> On Sun, Nov 3, 2013 at 8:37 PM, Alok Gandhi < >>>>>>>>>>>> [email protected]> wrote: >>>>>>>>>>>> >>>>>>>>>>>>> I am a little confused... >>>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>>> For sure, you cannot have multiple objects in the same .pc2 >>>>>>>>>>>>> file, the format simply doesn't support that. >>>>>>>>>>>>> >>>>>>>>>>>>> >>>>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> -- >>>>>>>>>>> >>>>>>>>>> >>>>>>>>>> >>>>>>>>> >>>>>>>> >>>>>>> >>>>>> >>>>>> >>>>>> -- >>>>>> >>>>> >>>>> >>>> >>>> >>>> -- >>>> >>> >>> >>> >>> -- >>> >> >> >> >> -- >> > > > > -- >
