Hi, I am the maintainer of GRENITS package. It has recently started to give build errors due to a function it uses from the package ggplot2. I've updated ggplot2 to the latest cran version (0.8.9) and it's dependencies but cannot replicate the errors. However, I found a newer version of ggplot2 (0.9.0) on crantastic that relies on the new package reshape2 (previously it used reshape). Could the problem be that ggplot2 was updated to the latest version (0.9.0), but reshape2 was not installed? How can I check the versions of the packages used for the build process? Thanks, Edward
[Bioc-devel] Package crashes with
7 messages · Ed, Laurent Gatto, Benilton Carvalho +1 more
Dear Edward, I believe the error is the result of ggplot2 v 0.8.9 depending on reshape while ggplot2 v 0.9.0 now imports reshape2. You now need to load reshape/reshape2 explicitly in your vignette prior to calling melt. Note that this change will have to be committed to devel and RELEASE, as ggplot2 has been updated on both branches. Hope this helps, Laurent On 2 February 2012 14:56, Edward Morrissey
<edward.morrissey at cancer.org.uk> wrote:
Hi, I am the maintainer of GRENITS package. It has recently started to give build errors due to a function it uses from the package ggplot2. I've updated ggplot2 to the latest cran version (0.8.9) and it's dependencies but cannot replicate the errors. However, I found a newer version of ggplot2 (0.9.0) on crantastic that relies on the new package reshape2 (previously it used reshape). Could the problem be that ggplot2 was updated to the latest version (0.9.0), but reshape2 was not installed? How can I check the versions of the packages used for the build process? Thanks, Edward
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
Hey Laurent, The major issue, IMHO, is the fact that internally ggplot2 has distinct dependencies conditional on its version. ggplot2 0.8.9 depends on reshape; ggplot2 0.9.0 depends on reshape2. And both reshape and reshape2 provide different implementations of the same method (see melt). If ggplot2 0.9.0 was available for every platform, the changes in GRENITS (and whatever other package that uses ggplot2) would be straightforward (but still require some extra work, given how the format of some outputs from reshape2 changed compared to reshape)... And it does not seem to me that these are changes that we expect to do in the middle of a cycle on the stable branch. One particular issue is the fact the Hadley mentioned that ggplot2 0.9.0 would be resubmitted to CRAN only in March ( http://goo.gl/p5rbq )... which makes this slightly harder to handle. Maybe we'd need to have one exception and simply force to have ggplot2 0.8.9 on the stable branch? Or just solve the issue that impedes ggplot2 0.9.0 to be built on every platform and deal with the fact that there would be major (internal) changes on the BioC release branch... b
On 2 February 2012 18:25, Laurent Gatto <laurent.gatto at gmail.com> wrote:
Dear Edward, I believe the error is the result of ggplot2 v 0.8.9 depending on reshape while ggplot2 v 0.9.0 now imports reshape2. You now need to load reshape/reshape2 explicitly in your vignette prior to calling melt. Note that this change will have to be committed to devel and RELEASE, as ggplot2 has been updated on both branches. Hope this helps, Laurent On 2 February 2012 14:56, Edward Morrissey <edward.morrissey at cancer.org.uk> wrote:
Hi, I am the maintainer of GRENITS package. It has recently started to give build errors due to a function it uses from the package ggplot2. I've updated ggplot2 to the latest cran version (0.8.9) and it's dependencies but cannot replicate the errors. However, I found a newer version of ggplot2 (0.9.0) on crantastic that relies on the new package reshape2 (previously it used reshape). Could the problem be that ggplot2 was updated to the latest version (0.9.0), but reshape2 was not installed? How can I check the versions of the packages used for the build process? Thanks, Edward
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
Dear Benilton,
On 2 February 2012 19:01, Benilton Carvalho <beniltoncarvalho at gmail.com> wrote:
Hey Laurent, The major issue, IMHO, is the fact that internally ggplot2 has distinct dependencies conditional on its version. ggplot2 0.8.9 depends on reshape; ggplot2 0.9.0 depends on reshape2. And both reshape and reshape2 provide different implementations of the same method (see melt).
Definitely, that does have broader implications. Thank you for pointing it out.
If ggplot2 0.9.0 was available for every platform, the changes in GRENITS (and whatever other package that uses ggplot2) would be straightforward (but still require some extra work, given how the format of some outputs from reshape2 changed compared to reshape)... And it does not seem to me that these are changes that we expect to do in the middle of a cycle on the stable branch. One particular issue is the fact the Hadley mentioned that ggplot2 0.9.0 would be resubmitted to CRAN only in March ( http://goo.gl/p5rbq )... which makes this slightly harder to handle. Maybe we'd need to have one exception and simply force to have ggplot2 0.8.9 on the stable branch? Or just solve the issue that impedes ggplot2 0.9.0 to be built on every platform and deal with the fact that there would be major (internal) changes on the BioC release branch...
I was surprised that 0.9.0 was installed on the stable branch, as it is not yet available on CRAN. Not sure what the best approach is, though, as maintainers have or are on their way to make the necessary changes in devel and release. More confusing, to me at least, is that devel and release packages are build/checked with ggplot2 0.9.0 but biocLite installs 0.8.9. Best wishes, Laurent
b On 2 February 2012 18:25, Laurent Gatto <laurent.gatto at gmail.com> wrote:
Dear Edward, I believe the error is the result of ggplot2 v 0.8.9 depending on reshape while ggplot2 v 0.9.0 now imports reshape2. You now need to load reshape/reshape2 explicitly in your vignette prior to calling melt. Note that this change will have to be committed to devel and RELEASE, as ggplot2 has been updated on both branches. Hope this helps, Laurent On 2 February 2012 14:56, Edward Morrissey <edward.morrissey at cancer.org.uk> wrote:
Hi, I am the maintainer of GRENITS package. It has recently started to give build errors due to a function it uses from the package ggplot2. I've updated ggplot2 to the latest cran version (0.8.9) and it's dependencies but cannot replicate the errors. However, I found a newer version of ggplot2 (0.9.0) on crantastic that relies on the new package reshape2 (previously it used reshape). Could the problem be that ggplot2 was updated to the latest version (0.9.0), but reshape2 was not installed? How can I check the versions of the packages used for the build process? Thanks, Edward
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
[ Laurent Gatto | slashhome.be ]
The issues you're observing are likely associated to the fact that CRAN provides 0.8.9 for Linux and Windows... but 0.9.0 for Mac. b
On 3 February 2012 14:30, Laurent Gatto <laurent.gatto at gmail.com> wrote:
Dear Benilton, On 2 February 2012 19:01, Benilton Carvalho <beniltoncarvalho at gmail.com> wrote:
Hey Laurent, The major issue, IMHO, is the fact that internally ggplot2 has distinct dependencies conditional on its version. ggplot2 0.8.9 depends on reshape; ggplot2 0.9.0 depends on reshape2. And both reshape and reshape2 provide different implementations of the same method (see melt).
Definitely, that does have broader implications. Thank you for pointing it out.
If ggplot2 0.9.0 was available for every platform, the changes in GRENITS (and whatever other package that uses ggplot2) would be straightforward (but still require some extra work, given how the format of some outputs from reshape2 changed compared to reshape)... And it does not seem to me that these are changes that we expect to do in the middle of a cycle on the stable branch. One particular issue is the fact the Hadley mentioned that ggplot2 0.9.0 would be resubmitted to CRAN only in March ( http://goo.gl/p5rbq )... which makes this slightly harder to handle. Maybe we'd need to have one exception and simply force to have ggplot2 0.8.9 on the stable branch? Or just solve the issue that impedes ggplot2 0.9.0 to be built on every platform and deal with the fact that there would be major (internal) changes on the BioC release branch...
I was surprised that 0.9.0 was installed on the stable branch, as it is not yet available on CRAN. Not sure what the best approach is, though, as maintainers have or are on their way to make the necessary changes in devel and release. More confusing, to me at least, is that devel and release packages are build/checked with ggplot2 0.9.0 but biocLite installs 0.8.9. Best wishes, Laurent
b On 2 February 2012 18:25, Laurent Gatto <laurent.gatto at gmail.com> wrote:
Dear Edward, I believe the error is the result of ggplot2 v 0.8.9 depending on reshape while ggplot2 v 0.9.0 now imports reshape2. You now need to load reshape/reshape2 explicitly in your vignette prior to calling melt. Note that this change will have to be committed to devel and RELEASE, as ggplot2 has been updated on both branches. Hope this helps, Laurent On 2 February 2012 14:56, Edward Morrissey <edward.morrissey at cancer.org.uk> wrote:
Hi, I am the maintainer of GRENITS package. It has recently started to give build errors due to a function it uses from the package ggplot2. I've updated ggplot2 to the latest cran version (0.8.9) and it's dependencies but cannot replicate the errors. However, I found a newer version of ggplot2 (0.9.0) on crantastic that relies on the new package reshape2 (previously it used reshape). Could the problem be that ggplot2 was updated to the latest version (0.9.0), but reshape2 was not installed? How can I check the versions of the packages used for the build process? Thanks, Edward
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
-- [ Laurent Gatto | slashhome.be ]
Hi all,
On Fri, Feb 3, 2012 at 6:30 AM, Laurent Gatto <laurent.gatto at gmail.com> wrote:
Dear Benilton, On 2 February 2012 19:01, Benilton Carvalho <beniltoncarvalho at gmail.com> wrote:
Hey Laurent, The major issue, IMHO, is the fact that internally ggplot2 has distinct dependencies conditional on its version. ggplot2 0.8.9 depends on reshape; ggplot2 0.9.0 depends on reshape2. And both reshape and reshape2 provide different implementations of the same method (see melt).
Definitely, that does have broader implications. Thank you for pointing it out.
If ggplot2 0.9.0 was available for every platform, the changes in GRENITS (and whatever other package that uses ggplot2) would be straightforward (but still require some extra work, given how the format of some outputs from reshape2 changed compared to reshape)... And it does not seem to me that these are changes that we expect to do in the middle of a cycle on the stable branch. One particular issue is the fact the Hadley mentioned that ggplot2 0.9.0 would be resubmitted to CRAN only in March ( http://goo.gl/p5rbq )... which makes this slightly harder to handle. Maybe we'd need to have one exception and simply force to have ggplot2 0.8.9 on the stable branch? Or just solve the issue that impedes ggplot2 0.9.0 to be built on every platform and deal with the fact that there would be major (internal) changes on the BioC release branch...
I was surprised that 0.9.0 was installed on the stable branch, as it is not yet available on CRAN. Not sure what the best approach is, though, as maintainers have or are on their way to make the necessary changes in devel and release. More confusing, to me at least, is that devel and release packages are build/checked with ggplot2 0.9.0 but biocLite installs 0.8.9.
It appears that ggplot2 0.9.0 was available through CRAN and then withdrawn to 0.8.9 (except on Mac, where 0.9.0 is still available). The Bioconductor build system installs from CRAN and tries to load the newest packages available. I removed ggplot 0.9.0 from the build machines which will trigger the installation of 0.8.9 in the next build cycle. I will ask Hadley about why 0.9.0 is available for Mac only. Thanks, Dan
Best wishes, Laurent
b On 2 February 2012 18:25, Laurent Gatto <laurent.gatto at gmail.com> wrote:
Dear Edward, I believe the error is the result of ggplot2 v 0.8.9 depending on reshape while ggplot2 v 0.9.0 now imports reshape2. You now need to load reshape/reshape2 explicitly in your vignette prior to calling melt. Note that this change will have to be committed to devel and RELEASE, as ggplot2 has been updated on both branches. Hope this helps, Laurent On 2 February 2012 14:56, Edward Morrissey <edward.morrissey at cancer.org.uk> wrote:
Hi, I am the maintainer of GRENITS package. It has recently started to give build errors due to a function it uses from the package ggplot2. I've updated ggplot2 to the latest cran version (0.8.9) and it's dependencies but cannot replicate the errors. However, I found a newer version of ggplot2 (0.9.0) on crantastic that relies on the new package reshape2 (previously it used reshape). Could the problem be that ggplot2 was updated to the latest version (0.9.0), but reshape2 was not installed? How can I check the versions of the packages used for the build process? Thanks, Edward
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
-- [ Laurent Gatto | slashhome.be ]
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
On Fri, Feb 3, 2012 at 9:42 AM, Dan Tenenbaum <dtenenba at fhcrc.org> wrote:
Hi all, On Fri, Feb 3, 2012 at 6:30 AM, Laurent Gatto <laurent.gatto at gmail.com> wrote:
Dear Benilton, On 2 February 2012 19:01, Benilton Carvalho <beniltoncarvalho at gmail.com> wrote:
Hey Laurent, The major issue, IMHO, is the fact that internally ggplot2 has distinct dependencies conditional on its version. ggplot2 0.8.9 depends on reshape; ggplot2 0.9.0 depends on reshape2. And both reshape and reshape2 provide different implementations of the same method (see melt).
Definitely, that does have broader implications. Thank you for pointing it out.
If ggplot2 0.9.0 was available for every platform, the changes in GRENITS (and whatever other package that uses ggplot2) would be straightforward (but still require some extra work, given how the format of some outputs from reshape2 changed compared to reshape)... And it does not seem to me that these are changes that we expect to do in the middle of a cycle on the stable branch. One particular issue is the fact the Hadley mentioned that ggplot2 0.9.0 would be resubmitted to CRAN only in March ( http://goo.gl/p5rbq )... which makes this slightly harder to handle. Maybe we'd need to have one exception and simply force to have ggplot2 0.8.9 on the stable branch? Or just solve the issue that impedes ggplot2 0.9.0 to be built on every platform and deal with the fact that there would be major (internal) changes on the BioC release branch...
I was surprised that 0.9.0 was installed on the stable branch, as it is not yet available on CRAN. Not sure what the best approach is, though, as maintainers have or are on their way to make the necessary changes in devel and release. More confusing, to me at least, is that devel and release packages are build/checked with ggplot2 0.9.0 but biocLite installs 0.8.9.
It appears that ggplot2 0.9.0 was available through CRAN and then withdrawn to 0.8.9 (except on Mac, where 0.9.0 is still available). The Bioconductor build system installs from CRAN and tries to load the newest packages available. I removed ggplot 0.9.0 from the build machines which will trigger the installation of 0.8.9 in the next build cycle. I will ask Hadley about why 0.9.0 is available for Mac only.
Hadley has asked CRAN to pull ggplot2 0.9.0 for Mac and it looks like it indeed has reverted to 0.8.9 on the CRAN master. When this propagates to our CRAN mirror, I'll make sure we have 0.8.9 on our Mac build machine so all are consistent. Hadley also asked me for the email addresses of all BioC maintainers who depend on ggplot2. I imagine he will be in touch with you with information about how to make the transition when the time comes. Thanks, Dan
Thanks, Dan
Best wishes, Laurent
b On 2 February 2012 18:25, Laurent Gatto <laurent.gatto at gmail.com> wrote:
Dear Edward, I believe the error is the result of ggplot2 v 0.8.9 depending on reshape while ggplot2 v 0.9.0 now imports reshape2. You now need to load reshape/reshape2 explicitly in your vignette prior to calling melt. Note that this change will have to be committed to devel and RELEASE, as ggplot2 has been updated on both branches. Hope this helps, Laurent On 2 February 2012 14:56, Edward Morrissey <edward.morrissey at cancer.org.uk> wrote:
Hi, I am the maintainer of GRENITS package. It has recently started to give build errors due to a function it uses from the package ggplot2. I've updated ggplot2 to the latest cran version (0.8.9) and it's dependencies but cannot replicate the errors. However, I found a newer version of ggplot2 (0.9.0) on crantastic that relies on the new package reshape2 (previously it used reshape). Could the problem be that ggplot2 was updated to the latest version (0.9.0), but reshape2 was not installed? How can I check the versions of the packages used for the build process? Thanks, Edward
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
-- [ Laurent Gatto | slashhome.be ]
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel