New version of Bio++ and a bug in Nhx class

[Yifei Huang]

Hi everyone,

I found that the most recent version downloaded from the git repository is very different from the previous stable versions. For example, a lot of libraries have different interfaces and several classes have been merged. Do you suggest to switch to the new version or stick to the stable version now? In addition, I found the Nhx class does not work anymore. It used to work in the git version I downloaded in Jan. 26, 2012, but does not work in the most recent git version and bio++ 2.0.3 (I used exactly the same tree file). The error information is following:

terminate called after throwing an instance of 'bpp::Exception'

  what():  NestedStringTokenizer (constructor). Unclosed block.

Aborted

I attached the example program and the tree I used. Any help is highly appreciated. Thank you very much!

Best,

Yifei

--
Yifei Huang
Department of Biology
McMaster University

[Julien Yann Dutheil]

Hi Yifei,

I have fixed the Nhx bug (at least on your file). There are indeed
some changes since the last stable version, particularly in bpp-phyl.
I think most of them are documented on the devel forum. It is really
up to you to decide on whether you want to shift to the development
version. 99% of the changes on the git version are expected to be on
the next stable version, as we try to keep experimental features on
dedicated branches so that the master branch remains as stable as
possible.

All the best,

J

> --
> You received this message because you are subscribed to the Google Groups
> "Bio++ Usage Help Forum" group.
> To post to this group, send email to biopp-he...@googlegroups.com.
> To unsubscribe from this group, send email to
> biopp-help-for...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>



--
Julien Y. Dutheil, Ph-D
0 (+49) 6421 178 986

§ Max Planck Institute for Terrestrial Microbiology
Department of Organismic Interactions
Marburg -- GERMANY

§ Intitute of Evolutionary Sciences - Montpellier
University of Montpellier 2 -- FRANCE

[Yifei Huang]

Hi Julien,

Thanks a lot for your reply. I decided to move to the new version of bio++. By the way, I downloaded the new git clone today. However, it seems the Nhx class is still buggy. It has the following problems:

1) In the tree file, there are 6 nodes. However, in the parsed tree there are only 3 nodes.

2) The properties cannot be extracted from the nodes using the registered names.

I attached my code for testing. In addition, I tried "make test" for bpp-phyl, but test_nhx and test_likelihood_nh failed.

Best,

Yifei

PS: In Phyl.all, files are missing for the following two lines:

#include "Phyl/NewLikelihood.all"

#include "Phyl/OldLikelihood.all" 

[Julien Yann Dutheil]

Hi Yifei,

Thanks a lot for your feedback! I realized that the dashboard was not
running the unit tests nightly, most likely because of the recent
server update, which explains why I did not see the pb with the
test_nhx before :s.
I have committed a fix for the Nhx parser, which should deal with more
special cases (the issue was the presence/absence of annotations at
the root node). Hope it will work now.
Concerning the test_likelihood_nh unit test, is it constantly failing
on your machine? This can be due to distinct rounding errors, as the
test perform well on my machine for instance...

As for point 2, the property tag is "G" in your example, not "count".
Maybe try with Nhx(false) to use the property names (see
http://biopp.univ-montp2.fr/apidoc/bpp-phyl/html/classbpp_1_1Nhx.html#a001fbb99e4204e08b904bd9f8a9d68f7).

All the best,

Julien.

[Yifei Huang]

Hi Julien,

Sorry for the late of reply. I downloaded the new version and it works very well. It seems the test_likelihood_nh unit test has no problem in the new version. I don't know why it was failed in the older version but it seems everything is good now. Thanks a lot for your help!

Best,

Yifei

On Wed, Nov 21, 2012 at 7:39 AM, Julien Yann Dutheil <jy.du...@gmail.com> wrote:

I have committed a fix for the Nhx parser, which should deal with more
special cases (the issue was the presence/absence of annotations at
the root node). Hope it will work now.
Concerning the test_likelihood_nh unit test, is it constantly failing
on your machine? This can be due to distinct rounding errors, as the
test perform well on my machine for instance...

[Yifei Huang]

Hi Julien,

Sorry to bother you again, but I found another small problem in the Nhx parser. I found the Nhx parser cannot read the branch lengths from the tree file. I attached an example which shows the branch lengths in the tree object are 0's instead of the values in the tree file.

Best,

Yifei 

[Julien Yann Dutheil]

Hi Yifei,

Sorry about that... I think you forgot to attach the example file
showing the bug? Would you mind posting it on the mantis bug tracker
(biopp.univ-montp2.fr/bug) ? I think it is a better place for
reporting this type of issue. Thanks a lot for your patience and
careful debugging!

Cheers,

J.

[Yifei Huang]

Hi Julien,

Thanks for your reply. I submitted the problem to the mantis bug tracker.

Best,

Yifei

[Julien Yann Dutheil]

Hi Yifei,

Cannot remember if I said it already... but I have hopefully corrected
this issue... sorry for the late notice :s

J.

[Yifei Huang]

Thanks a lot, Julien. I tried the new code and it seems good.

Best,

Yifei