问题
According to this link, SQL Server machine learning service's highest R version is 3.3. Does anyone know when one can upgrade to version 3.5 or later?
Or is my only option to switch to sql server 2019?
Thanks.
回答1:
Well cs0815, let's take a look.
According to your post, you tagged this with sql-server-2017, so I'll be basing this on that version.
After reading through your link and this one, I was thinking there might be two options:
- Hack a new version of R into the R-Services module of SQL server.
- Setup an R-Client with version 3.5
However, it looks like R-Client is for a different use case than you might need so we'll focus on an attempt to replace R in R-Services with a higher version directly.
After installing a new instance of 2017 with the following features:
R Services natively installs to:
"C:\Program Files\Microsoft SQL Server\MSSQL14.MSSQLSERVER\R_SERVICES"
Using WinDiff, I pulled the following comparisons vs CRAN R 3.5.3 and Microsoft Open R 3.5.3 and not surprisingly the Microsoft Open R is much closer.
Full diffs: R_Services vs CRAN R R_Services vs MS OPEN R
Personally, I'm not at the level to understand all the differences in R Core / runtime between R 3.3 (R_Services) and R 3.5.3 so of main interest to me is the library diff.
What exists in R_Services which does not in Open R is what I'm assuming to be the core of what makes R_Services work with SQL Server so that will be where we need to fill in those gaps with our "replacement" R_Services but with v3.5.3.
Libraries which are completely missing - not just added docs or some small difference:
- compatibilityapi
- dorsr
- microsoftml
- mrsdeploy
- olapr
- revopemar
- revoscaler
- revotreeview
- sqlrutils
And that information leads to us to the Microsoft MLS Docs "R Packages" resource:
https://docs.microsoft.com/en-us/machine-learning-server/r-reference/introducing-r-server-r-package-reference
Which, no, it looks like those packages are not publicly available through MRAN, CRAN, or even directly via devtools from github.
Still, searching brought me onto this possibility: https://docs.microsoft.com/en-us/sql/advanced-analytics/install/upgrade-r-and-python?view=sql-server-2017#offline-binding-no-internet-access
Download after signing up with Visual Studio Dev Services: https://docs.microsoft.com/en-us/machine-learning-server/install/machine-learning-server-windows-install#download-machine-learning-server-installer
So - binding this version of MLS to our existing instance:
After that installation is fully complete, let's check and see what's available on our server!
So first things first, this being a new instance of SQL Server - Set to enable external scripts:
sp_configure 'external scripts enabled', 1;
RECONFIGURE WITH OVERRIDE;
And then to bring this soup to nuts:
EXECUTE sp_execute_external_script @language = N'R'
, @script = N'
print(R.Version())
';
output below:
$language
[1] "R"
$version.string
[1] "R version 3.5.2 (2018-12-20)"
$nickname
[1] "Eggshell Igloo"
TLDR: Yes - download the MLS 9.4.7 packages from Visual Studio Dev Services and install on server host.
来源：https://stackoverflow.com/questions/59364167/sql-server-machine-learning-services-r-version-3-5