Camelia, the Perl 6 bug

IRC log for #cdk, 2008-01-17

| Channels | #cdk index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
05:25 jbrefort joined #cdk
05:37 masak joined #cdk
07:01 CIA-17 cdk: egonw * r9916 /branches/egonw/charge/src/org/ope​nscience/cdk/qsar/descriptors/atom​ic/PartialPiChargeDescriptor.java: Fixed wrapping of charges
07:03 CIA-17 cdk: egonw * r9917 /branches/egonw/charge/src/org/opensc​ience/cdk/test/qsar/descriptors/atomi​c/PartialPiChargeDescriptorTest.java: More info when failing: added fail messages
07:08 CIA-17 cdk: egonw * r9918 /trunk/cdk/src/org/openscience/cd​k/config/data/cdk_atomtypes.xml: A bit more comment... should really move to OWL soon
07:14 CIA-17 cdk: egonw * r9919 /trunk/cdk/src/org/openscience/cdk/test/ (2 files in 2 dirs): Added unit tests for bug report 1872969
07:37 rojasm joined #cdk
07:37 rojasm moin
07:54 CIA-17 cdk: miguelrojasch * r9920 /branches/miguelrojasch/mf/src/org/openscien​ce/cdk/test/modulesuites/MformulaTests.java: Updated to JUnit4.
07:56 CIA-17 cdk: miguelrojasch * r9921 /branches/miguelrojasch/mf/src/org/openscienc​e/cdk/test/formula/MolecularFormulaTest.java: FIXME: strange situation in the testRemoveIsotope_Int. Don't give always the same result.
07:57 Annzi joined #cdk
08:34 egonw_ joined #cdk
08:35 egonw_ moin
08:59 rojasm moin, egonw
08:59 rojasm For the next Frieday is perfect
09:05 egonw_ ok, good
09:05 egonw_ will try to be there as early as possible
09:05 egonw_ really would like to fix a lot of those CDK bugs in charges, IP, reaction...
09:05 egonw_ that seems to be the main show stopper for trunk/
09:06 rojasm perfect
09:06 rojasm But I have to leave 5 afternoon
09:07 rojasm Do you think it will be still sufficient
09:07 rojasm I sent now
09:07 rojasm you
09:08 egonw_ yeah, should be fine...
09:08 egonw_ will leave really early :)
09:09 egonw_ I will be there at 9 o'clock
09:09 egonw_ if there are no delays
09:09 rojasm Done
09:10 steinbeck joined #cdk
09:10 rojasm Perfect
09:10 rojasm hi steinbeck
09:10 rojasm morning
09:11 egonw_ hi steinbeck
09:11 jonalv joined #cdk
09:11 steinbeck moin
09:25 jbrefort joined #cdk
09:54 rojasm egonw: that is great. I will be of course there
10:00 rojasm egonw: that sounds good
10:01 rojasm you can not see my answer in privat chat, right?
10:11 egonw_ no, I can't
10:11 rojasm :(
10:11 rojasm ok
10:11 egonw_ still haven't recovered your password, I guess? :)
10:11 CIA-17 cdk: miguelrojasch * r9922 /branches/miguelrojasch/mf/src​/org/openscience/cdk/formula/ (IMolecularFormula.java MolecularFormula.java): remove method and test which get the isotope from position. HashMap doesn't recover the same object in the same position as was stored.
10:13 rojasm egonw: you put in some bugs mail to use Hashtable instead of Map. Why not HashMap?
10:17 rojasm Now I see why in my tests I don't obtained the same results
10:17 rojasm The HashMap class is roughly equivalent to Hashtable, except that HashMap unsynchronized and permits nulls. (HashMap allows null values as key and value whereas Hashtable doesnt allow). HashMap does not guarantee that the order of the map will remain constant over time. HashMap is non synchronized and Hashtable is synchronized.
10:19 egonw_ right
10:20 egonw_ but Hashtable vs Map is about the API using an interface
10:20 egonw_ not a Class
10:20 rojasm ok
10:21 egonw_ not sure how to ensure the ordering of items
10:23 rojasm mmm. But nevertherless we supose that in Hashtable the items will remain allways in the same order, right?
10:34 CIA-17 cdk: miguelrojasch * r9923 /branches/miguelrojasch/mf​/src/org/openscience/cdk/ (18 files in 2 dirs): name of the test changed according to the name of the method. Easier looking the realation in Junit4
11:25 * egonw_ is relocating to PRI
11:35 egonw moin (at PRI now)
11:56 steinbeck joined #cdk
12:02 masak joined #cdk
12:05 jonalv rojasm, A hashtable guarantees absolutely nothing about the order of it's items
12:07 rojasm I checked it. That is true
12:07 rojasm I have still the same problem
12:07 jonalv there is something called LinkedHashMap...
12:08 rojasm I think you commented me already mounths ago.
12:08 jonalv yes I am also having flashbacks...
12:08 jonalv or dejavue
12:09 jonalv So why didn't it work?
12:09 rojasm but I don't know if makes sense to have the method
12:09 rojasm I have a class IMolecularFormula which can added IISotopes
12:09 rojasm and one of the method was getIsotope(int position)
12:10 jonalv okey?
12:10 rojasm the Isotopes were stored in HashMap
12:10 jonalv heeh
12:10 jonalv Then you could use a LinedHashMap if you need to keep the method
12:10 jonalv or?
12:10 rojasm There is not code using this method. At least for me
12:11 jonalv okey well I don't know how wildly used that API is so I am not the right persopn to ask wether the method can be just removed
12:12 rojasm nobody because is still in my branches :)
12:12 rojasm also I don't think it will supose one problem
12:12 jonalv k
12:13 jonalv sounds like a design decission
12:14 rojasm But I don't understand why hashmap and hashtable don't mantain the order
12:14 jonalv why?
12:15 rojasm mantain == keep
12:15 rojasm sorry
12:15 jonalv hm it's about how the hashCode is implemented
12:16 jonalv THey might very well keep the order but they don't guarantee it so you shouldn't count ont it. Someone might do some optimisation somewhere that breaks it
12:24 masak jonalv: not really
12:24 masak it's more like you trade order for speed of access
12:24 masak because you can't have both
12:24 jonalv oh?
12:24 masak yes.
12:24 jonalv I haven't thought about ti that much
12:24 masak retrieving is O(1) in the best case
12:25 jonalv yes..
12:25 masak and, if you choose your hashcode ideally, in all cases
12:25 jonalv yes...
12:25 masak contrast a linear list, where retrieving is O(n), but you have perfect order
12:25 jonalv but I think the question was aboout retaining the order
12:26 masak if by order you mean 'insertion order' then LinkedHashMap is the answer
12:27 jonalv yes but I think the idea was that each time you iterate over the elemetnts in the map you get them in the same order
12:27 jonalv no necessarily the order you inserted them in
12:27 masak jonalv: but any order will do?
12:27 jonalv All I said was that there are no such guarantees
12:27 masak jonalv: well...
12:28 masak technically, as long as you don't insert or delete anything...
12:28 masak but it feels wrong to rely on, that's true
12:28 jonalv yes but don't go there is all I am saying
12:29 masak ACK
12:29 masak :)
12:30 egonw I think the HashMap uses a tree structure internally...
12:30 egonw to optimize lookup...
12:30 egonw for best performance, such a thing is balanced
12:30 egonw so, you balance the things every now and then
12:31 egonw but that will mess up the order
12:31 jonalv you should program againt the interface Map and not bother with how the implemmenation is done I think...
12:31 jonalv yes if that's the case what I fist said was true
12:33 jonalv They uses what works best and they are free to change between versions so I don't see where this discussion is going...
12:34 egonw agreed
12:34 masak egonw: unless you know something I don't about HashMap, it's not implemented as a tree. however, the main point still stands: don't much with internals
12:34 egonw rojasm: and you need to look for another machanism to match old objects in the Map with the new ones
12:35 rojasm I don't think it is necessary the option to know what order was introduce the items, isn't it?
12:36 rojasm I thought it could be nice. But better to have fast method
12:36 jonalv how many item are we talking about here?
12:37 rojasm you mean how many entries could have the HashMap?
12:37 jonalv yes
12:38 rojasm It defines the Isotopes which can contain a molecular formula
12:38 jonalv okey...
12:39 jonalv My point is that if they are not *very* many the speed of the datastructure used might not be so important
12:41 rojasm I think I will leave as HashMap at the moment!
12:41 rojasm thanks all for helping
12:42 jonalv ?!
12:42 jonalv all this for nothing?
12:42 rojasm ?
12:43 jonalv oh never mind...
13:04 thomas_ku joined #cdk
13:05 thomas_ku moin
13:06 egonw hi thomas_ku
13:07 thomas_ku egonw: sorry hadn't got the time to look into your configuration thingy up to now
13:07 thomas_ku will to it later in the afternoon
13:07 egonw no problem
13:08 egonw thanx
13:08 egonw I haven't had more time for it either...
13:08 egonw any input/ideas is most welcome...
13:09 thomas_ku within your mail there were the cdk and cdk-taverna jars missing
13:09 thomas_ku was it because of the large amount of data?
13:13 Annzi joined #cdk
13:18 Annzi joined #cdk
13:37 olas joined #cdk
13:47 thomas_ku egonw: how do I update the ant within eclipse? I have a local installation which works on the shell but what do I need to do within this beautifull IDE?
13:50 egonw no idea
13:51 egonw can't remember manually installing ant in eclipse
13:51 thomas_ku ok... than I will look
13:57 dleidert joined #cdk
14:16 thomas_ku egonw: did you removed the jchempaint from the pom file?
14:20 egonw yes
14:20 egonw but maybe not in SVN
14:21 thomas_ku oh yes it is remove within svn
14:21 thomas_ku did you commented out the jcp worker too?
14:24 egonw I might not have done that...
14:24 egonw let me try
14:25 thomas_ku please also remove the jcp action class than
15:24 rojasm egonw: around?
15:28 egonw yes
15:29 rojasm I  have a question about MF!
15:29 rojasm I would like to put into a reaction a IMF
15:29 rojasm but for that it should be a instance of AtomContainer
15:29 rojasm and now It is not
15:30 egonw a reaction assumes some connectivity...
15:30 egonw a reaction does not have that
15:30 egonw ummm
15:30 egonw a MF does not have that
15:31 rojasm But you can define a reaction without assumes connectivity
15:32 egonw yes, but that's not the scope of IReaction
15:32 rojasm that is true
15:34 rojasm But the same idea that you define an Atom as X why not a Molecule as MYMOLECULE000 and why not reaction between A + B => C
15:34 egonw SMIRKS?
15:35 egonw why not a simple String ?
15:35 egonw like "A + B > C" ?
15:35 rojasm not exactly. I mean we can create an empty IAtomContainer and put into a reaction
15:35 egonw yes, that's is possible too
15:35 egonw with a TITLE set
15:36 rojasm But not with a IMolecularFormula for example!
15:36 egonw true
15:36 egonw why do you need that?
15:36 egonw would IMF not be an overkill?
15:36 rojasm because I think there is a step because chemObject and IAtomContainer
15:37 rojasm reaction ist between proteins as well
15:37 rojasm and not only molecules
15:37 rojasm ok, it was only a proposal
15:38 egonw don't give up...
15:38 egonw I'
15:38 egonw m not convinced yet...
15:38 egonw but not dead against extensions...
15:38 rojasm :)
15:38 egonw if there is a good use
15:38 egonw what algorithm would make use of it?
15:40 rojasm I am thinking on defining a reaction with MF
15:40 rojasm because now the spectral tree definition contains only information about that
15:40 rojasm of course it should contain a parent ion
15:41 rojasm other the chemical structure but we don't know.
15:41 egonw so, the question is... why do you need such reactions expressed in IChemObjects?
15:41 rojasm But may be somebody would be interested also in reaction of proteins or at least the definition
15:42 rojasm Because that mean more flexible to use reactions
15:43 rojasm reactions don't check if it is possible but only the definition take acount some thing gives some others
15:43 rojasm that is because I think may be should be better to pot products as IChemObjects
15:44 rojasm products and reactants
15:44 egonw I suggest we wait until algorithms that act on such a representation surface
15:45 rojasm I can show you wenn you come here
15:48 egonw excellent
16:02 steinbeck joined #cdk
16:23 rojasm left #cdk
16:32 Annzi left #cdk
16:47 * egonw is away for basketbal
17:05 olas left #cdk
17:25 CIA-18 joined #cdk
20:07 carsten joined #cdk
20:27 edrin joined #cdk
22:32 CIA-18 bioclipse: ospjuth * r4707 /branches/bioclipse2/core/​plugins/net.bioclipse.ui/ (bioclipse.product src/net/bioclipse/ui/Activator.java): Updated product and Activator for ui plugins.
22:32 CIA-18 bioclipse: ospjuth * r4708 /branches/bioclipse2/chemoinformatics/net.biocli​pse.qsar/src/net/bioclipse/qsar/Activator.java: Activator needs its own getDefault()
23:19 CIA-18 bioclipse: ospjuth * r4709 /branches/bioclipse2/bioinformati​cs/plugins/net.bioclipse.biojava/ (.classpath META-INF/MANIFEST.MF jars/biojava.jar): Temoved old jar and updated exported pkgs
23:20 CIA-18 bioclipse: ospjuth * r4710 /branches/bioclipse2/core/plugins/ne​t.bioclipse.core/src/net/bioclipse/c​ore/domain/ISequenceCollection.java: Added interface for multiple sequences
23:24 CIA-18 bioclipse: ospjuth * r4711 /branches/bioclipse2/bioinformatic​s/plugins/net.bioclipse.biojava.bu​siness/src/net/bioclipse/biojava/ (4 files in 2 dirs): Added BiojavaSequenceCollection for containing multiple sequences
23:26 CIA-18 bioclipse: ospjuth * r4712 /branches/bioclipse2/bioinformatics/plugins/​net.bioclipse.biojava.business.tests/tests/ (4 files in 2 dirs): Added test for containing multiple sequences

| Channels | #cdk index | Today | | Search | Google Search | Plain-Text | summary