Unable to do SciDB R operation


#1

SciDB-R should support scalar multiplication according to the documentation -https://cran.r-project.org/web/packages/scidb/vignettes/scidb.pdf
But I am getting error messages.

X <- matrix(runif(16),4)
A <- as.scidb(X)
2*A
Error in 2 * A : non-numeric argument to binary operator

Further I checked I am not able to any operations like
A+A
crossprod(A,A)

What am I doing wrong?


#2

Hi theabhash

I’m having the same problem than you.

Did you get it fixed?

Thanks!


#3

@vdebuen

The scidbR package was updated and the operations mentioned here are not supported anymore. The current CRAN vignette should not have any references to these operations so maybe you are using an older copy.

Please make sure you are using the latest version of scidbR either from CRAN, or better still from Github:

devtools::install_github('paradigm4/scidbr')

#4

Thanks you Kriti for your answer.

I have tried both the CRAN and github versions and the problem is the same, as these operations are no longer supported, in fact I see that the only operation supported is the calculation of the covariance matrix, which in fact needs internally both the product and the sum of matrices, so I find this change of development strategy very strange.

The truth is that the integration with R at the level of linear algebra operators, as described in https://www.paradigm4.com/try_scidb/scidb-r/, was what interested me most about SciDb, if not the only thing.

It seems to me that it is a great loss to eliminate the operators of linear algebra in R, especially being something that apparently was working well.

I had planned to use the professional version of SciDB in the Azure cloud in a project that will start in a few days, but I’m afraid it’s not going to be possible, unless in the version for companies these operators are still maintained, and there are guarantees that it will continue to be so in the future.

I’m very sorry, but if not, I’m no longer interested in this product because it was an indispensable condition to be able to use it in my work environment.

Best regards

Translated with www.DeepL.com/Translator


#5

@vdebuen

Thanks for your response. Let me add a few thoughts.

Firstly, you should be able to install the older version of scidbR using

devtools::install_github('paradigm4/scidbr', ref='discontinued_master')

OK, moving on to other things. You said:

I understand your point. The demo does show the power of the older scidbR design where one could code directly in R and the operations would happen under the hood in scidb. Now while the demo did work (and it should continue to work with the above referenced installation method), we found that users fell into various edge cases with this design i.e. it was being difficult to allow seamless integration of SciDB operations with all sorts of R code.

That is why we moved to the current design – asking the user to explicitly write the scidb operations that they want to do. This requires the user to learn more on scidb operators (translating into a steeper learning curve) but this also greatly reduced the errors due to edge cases (translating into easier maintainability for open software like scidbR).

Finally, I would like to mention that we recently completed a custom project in which multiple matrix operations can be carried out using REST API (i.e. allowing interconnectivity to any programming environment). The project provided a high level matrix programming API that does not require the user to learn any of scidb’s internals. At this point, I believe your needs can be met with such a paid / collaborative project. If you are interested, please contact us at https://www.paradigm4.com/about/contact/. However if that is not feasible for you, we (at P4) appreciate your feedback and will reach out to you if the situation changes on our end.