Re: CI for mapdamage (Was: Outreachy 2017 (Round 14))
Hi Nadiya, On Wed, Mar 29, 2017 at 06:36:41PM -0700, Nadiya Sitdykova wrote: > I uncommented all tests in run-unit-test and now they work perfectly fine > so I closed the bug #848326 inside changelog > Also I added some info to README.test, which I think could be useful for > users Thanks for your fine work. I've uploaded the package and opened a bug against release.debian.org to unblock the testing migration to make sure users of next stable release will profit from these changes. I'm really happy that we have spotted and fixed a package that was basically non-functional before. Kind regards Andreas. -- http://fam-tille.de
Re: CI for mapdamage (Was: Outreachy 2017 (Round 14))
Hello Andreas, I uncommented all tests in run-unit-test and now they work perfectly fine so I closed the bug #848326 inside changelog Also I added some info to README.test, which I think could be useful for users Regards, Nadiya On Wed, Mar 29, 2017 at 2:48 PM, Nadiya Sitdykova wrote: > Hello Andreas, > > >Did you do some investigation what dependency providing liblapack.so and > >libblas.so are needed (see my mail from yesterday)? > > Yes, I tried to run the test with different packages installed, and here > are results: > > packages| result| user > time > > -- > libatlas-base-dev | Successful run | 11m24.204s > libopenblas-dev | Successful run | 10m31.860s > libblas-dev+liblapack-dev | Successful run | 10m34.368s > > It seems that libatlas-base-dev is less effective, but I'm not sure if we > should exclude it from Depends, so I left if for now > > On Tue, Mar 28, 2017 at 11:29 PM, Andreas Tille wrote: > >> [Changed subject to split up to topics in one thread] >> >> Hi Nadiya, >> >> On Tue, Mar 28, 2017 at 10:23:00PM -0700, Nadiya Sitdykova wrote: >> > 1) I read the paper about mapdamage and figured out that with given >> pe.sam >> > file the reference TGCGA wouldn't make sense as the particular type >> of >> > substitutions is specific for ancient DNA damages. So further for tests >> I >> > used reference CGATACGA -- which contains all types of nucleotides and >> make >> > sense regarding specifics of DNA damage. >> > >> > 2) But even after fixing (1) I still got an error (log attached). The >> > problem is that on given dataset from pe_test mapdamage produces data >> with >> > NaNs and then tries to compute quantiles on this data. I think that on >> real >> > date NaNs also could be produced, so I treated this issue as a bug and >> made >> > a patch which changes the way how quantiles are computed --- now it >> removes >> > NaNs [1] >> > >> > 3) I wrote about all this to upstream as a comments to earlier created >> > issue and asked if it's really was a bug in their opinion >> > >> > 4) FInally, move forward to fixing mapdamage test suite >> >> Pretty good job. That's exactly the contribution I'm envisioning for the >> Outreachy project. I've just added the URL of the upstream issue to the >> patch. >> >> > 11: stop("missing values and NaN's not allowed if 'na.rm' is FALSE") >> > 10: quantile.default(newX[, i], ...) >> > 9: FUN(newX[, i], ...) >> > 8: apply(X, 1, quantile, c(0.025)) >> > 7: data.frame(x = 1:nrow(da), pos = da[, "Pos"], mea = apply(X, >> >1, mean), med = apply(X, 1, median), loCI = apply(X, 1, quantile, >> >c(0.025)), hiCI = apply(X, 1, quantile, c(0.975))) >> > 6: calcSampleStats(da, CTs) >> > 5: postPredCheck(dat, mcmcOut) >> > 4: eval(expr, envir, enclos) >> > 3: eval(ei, envir) >> > 2: withVisible(eval(ei, envir)) >> > 1: source(paste(path_to_mapDamage_stats, "main.R", sep = "")) >> > The Bayesian statistics program failed to finish >> > Traceback (most recent call last): >> > File "/usr/local/bin/mapDamage", line 382, in >> > sys.exit(main()) >> > File "/usr/local/bin/mapDamage", line 365, in main >> > mapdamage.rscript.run_stats(options) >> > File "/usr/local/lib/python2.7/dist-packages/mapdamage/rscript.py", >> line 144, in run_stats >> > raise e >> > subprocess.CalledProcessError: Command '['Rscript', >> '/usr/local/lib/python2.7/dist-packages/mapdamage/Rscripts/stats/runGeneral.R', >> '--args', '30', '1', '10', '5', '0', '0', '1', '1', '0', '0', '1', >> '12', 'results_pe/', '/usr/local/lib/python2.7/dist >> -packages/mapdamage/Rscripts/stats/', 'results_pe/Stats_out', '0', '0', >> '0', 'results_pe/acgt_ratio.csv', '0', '0']' returned non-zero exit status 1 >> >> I'd recommend to manually run this script. >> >> Did you do some investigation what dependency providing liblapack.so and >> libblas.so are needed (see my mail from yesterday)? >> >> Kind regards >> >> Andreas. >> >> -- >> http://fam-tille.de >> >> >
Re: CI for mapdamage (Was: Outreachy 2017 (Round 14))
Hello Andreas, >Did you do some investigation what dependency providing liblapack.so and >libblas.so are needed (see my mail from yesterday)? Yes, I tried to run the test with different packages installed, and here are results: packages| result| user time -- libatlas-base-dev | Successful run | 11m24.204s libopenblas-dev | Successful run | 10m31.860s libblas-dev+liblapack-dev | Successful run | 10m34.368s It seems that libatlas-base-dev is less effective, but I'm not sure if we should exclude it from Depends, so I left if for now On Tue, Mar 28, 2017 at 11:29 PM, Andreas Tille wrote: > [Changed subject to split up to topics in one thread] > > Hi Nadiya, > > On Tue, Mar 28, 2017 at 10:23:00PM -0700, Nadiya Sitdykova wrote: > > 1) I read the paper about mapdamage and figured out that with given > pe.sam > > file the reference TGCGA wouldn't make sense as the particular type > of > > substitutions is specific for ancient DNA damages. So further for tests I > > used reference CGATACGA -- which contains all types of nucleotides and > make > > sense regarding specifics of DNA damage. > > > > 2) But even after fixing (1) I still got an error (log attached). The > > problem is that on given dataset from pe_test mapdamage produces data > with > > NaNs and then tries to compute quantiles on this data. I think that on > real > > date NaNs also could be produced, so I treated this issue as a bug and > made > > a patch which changes the way how quantiles are computed --- now it > removes > > NaNs [1] > > > > 3) I wrote about all this to upstream as a comments to earlier created > > issue and asked if it's really was a bug in their opinion > > > > 4) FInally, move forward to fixing mapdamage test suite > > Pretty good job. That's exactly the contribution I'm envisioning for the > Outreachy project. I've just added the URL of the upstream issue to the > patch. > > > 11: stop("missing values and NaN's not allowed if 'na.rm' is FALSE") > > 10: quantile.default(newX[, i], ...) > > 9: FUN(newX[, i], ...) > > 8: apply(X, 1, quantile, c(0.025)) > > 7: data.frame(x = 1:nrow(da), pos = da[, "Pos"], mea = apply(X, > >1, mean), med = apply(X, 1, median), loCI = apply(X, 1, quantile, > >c(0.025)), hiCI = apply(X, 1, quantile, c(0.975))) > > 6: calcSampleStats(da, CTs) > > 5: postPredCheck(dat, mcmcOut) > > 4: eval(expr, envir, enclos) > > 3: eval(ei, envir) > > 2: withVisible(eval(ei, envir)) > > 1: source(paste(path_to_mapDamage_stats, "main.R", sep = "")) > > The Bayesian statistics program failed to finish > > Traceback (most recent call last): > > File "/usr/local/bin/mapDamage", line 382, in > > sys.exit(main()) > > File "/usr/local/bin/mapDamage", line 365, in main > > mapdamage.rscript.run_stats(options) > > File "/usr/local/lib/python2.7/dist-packages/mapdamage/rscript.py", > line 144, in run_stats > > raise e > > subprocess.CalledProcessError: Command '['Rscript', > '/usr/local/lib/python2.7/dist-packages/mapdamage/Rscripts/stats/runGeneral.R', > '--args', '30', '1', '10', '5', '0', '0', '1', '1', '0', '0', '1', > '12', 'results_pe/', > '/usr/local/lib/python2.7/dist-packages/mapdamage/Rscripts/stats/', > 'results_pe/Stats_out', '0', '0', '0', 'results_pe/acgt_ratio.csv', '0', > '0']' returned non-zero exit status 1 > > I'd recommend to manually run this script. > > Did you do some investigation what dependency providing liblapack.so and > libblas.so are needed (see my mail from yesterday)? > > Kind regards > > Andreas. > > -- > http://fam-tille.de > >