memory not mapped in get_seurat #809
Comments
|
Glad the tip helped -- we are in the process to sort this out / smoth this a little but we are currently focussing on getting a first official 1.5.0 release of tiledb-soma out after which this should get become easier, (CC @johnkerl). |
|
Thanks Dirk, appreciate your help (all all your great Rcpp work!) So you think this is related? I understand. Any about a workaround for the current situation (downgrading tiledbsoma, couldn't find any older binary), or is this release planned soon anyway? How does it work on your machine? |
|
Yes related in the sense that compiled parts (and more so with C++ than with C) need to be aliged / can create issues when mimatched builds happen. I am more on the dev side here than the user side so I compile it all locally and it all works. But we have seen things similar to what you describe in our CI when using pre-built parts so we know about it, and as I mentioned, planned to put some improvements in. But first a full 1.5.0 release. |
|
@mihem While we're pushing out the 1.5.0 release, two possible temporary workarounds would be to:
Sorry for the inconvenience! |
|
@mlin thanks for you support. I would prefer upgrading to downgrading ;). But 1. leads to the same problem |
|
@eddelbuettel re @mihem's difficulties building from source...seems like something to do with |
|
@mlin I presume that is "theoretically possible" as @mihem It all works for me. You could, as @mlin suggests, unset |
|
thanks for your responses. Yes, I am on Ubuntu jammy 22.04 (also see first post environment). Sorry, maybe I mistunderstood you, but as explained in the first post your tip with installing |
|
Ok, sorry. The codespace mention was an illustration (possibly poorly worded) showing that everything works (well: worked when I wrote the blog post as an illustration/set this up in a Docker environment where all components come from consistently-built Ubuntu 22.04 LTS environment, namely my r-universe which has the 'unreleased' pre-1.5.0 release current tiledbsoma snapshot as a binary when you follow the link about to install them with the specific repo= argument to We are sadly in a bit of pre-release state but our CI also runs the tiledb (R package), tiledbsoma, Seurat, ... packages as Ubuntu 22.04 binaries (via r2u). I think CELLxGENE may just 'fit' with them but I am currently traveling and can't fully tests this for you. |
|
Detailed log below but I just
So in short we should still be able to get you set up. This is your script executed line by line in Emacs/ESS and output copied from the Emacs buffer. > library(cellxgene.census)
> census <- open_soma()
The stable Census release is currently 2023-07-25. Specify census_version = "2023-07-25" in future calls to open_soma() to ensure data consistency.
> organism <- "Homo sapiens"
> gene_filter <- "feature_id %in% c('ENSG00000107317', 'ENSG00000106034')"
> cell_filter <- "cell_type == 'sympathetic neuron'"
> cell_columns <- c("assay", "cell_type", "tissue", "tissue_general", "suspension_type", "disease")
> seurat_obj <- get_seurat(
+ census = census,
+ organism = organism,
+ var_value_filter = gene_filter,
+ obs_value_filter = cell_filter,
+ obs_column_names = cell_columns
!+ )
The legacy packages maptools, rgdal, and rgeos, underpinning the sp package,
which was just loaded, were retired in October 2023.
Please refer to R-spatial evolution reports for details, especially
https://r-spatial.org/r/2023/05/15/evolution4.html.
It may be desirable to make the sf package available;
package maintainers should consider adding sf to Suggests:.
Attaching SeuratObject
> |
|
Hello @mihem -- to repro, I tried using as a clean slate I was able to do the following: As noted above we're in a bit of flux as our 1.5.0 is almost release but not quite yet Also, this worked well for me -- but I don't have your particular ccache build error -- I can look more deeply into that tomorrow |
|
Wow, three experts working on this issue. I am little confused, why do you start with installing So if I start running your code,
When I start with installing |
@mihem the issue is that R doesn't allow upper-bound pins, and tiledb-r 0.21 has appeared since tiledbsoma 1.4.3 (which needs tiledb-r 0.20). So we indeed install tiledbsoma, then update the tiledb-r package to be 0.20. (This situation will disappear once we release 1.5.0 (which uses tiledb-r 0.21), and that 1.5.0 release will be soon. The current situation admits a non-ideal workaround.)
This checks out with what I saw above namely:
This is a bit odd: and I will indeed look into this more deeply today. My apologies for the inconvenience, and thank you! |
|
@mihem can you please share your |
|
sure, cmake version 3.22.1 |
|
@mihem I also have cmake version 3.22.1 on I believe the sole issue right now is why on your system cmake is detecting |
|
(I do not think (I suspect different versions of under Ubuntu 23.04 and |
|
@mihem with regard to https://ccache.dev/manual/4.8.3.html#_run_modes can we please check on your system: Can you please run all six of these and send me all their outputs? Thank you! |
|
@mihem also |
|
@johnkerl Recall that the build is started from R and in a shell spawned from R so settings as this (here from my box) may also come in when R is calling out. Hence eg $ R CMD config CXX
ccache g++
$ R CMD config CXX17
ccache g++
$
$ R CMD config CC
ccache gcc
$ which under normal settings gets back to the ones in the $PATH, |
and Dirks ideas |
|
@mihem aha, I have indeed |
|
@mihem If you have a It's a bit confusing but I think CXX is more or less ignore if C++17 is selected as compilation standard (as we do) Ditto @johnkerl You need (At least I think I have this right. Details in Writiing R Extensions as always.) |
|
@mihem can you please show us your |
|
@mihem I think John was right all along -- while I am a huge fanboy of PS It also no longer is an issue for us in dev all the |
|
Intersting. I then added a line for CXX17 as Dirk suggested (not sure if FLAGS were needed Now it doesn't stop with 56% but at 88% |
|
Could you humour us and please remove the setting for |
|
Ah, I am sorry, I didn't get that before. Yes, now tiledbsoma is compiling successfully. However, unfortunately i still canno access any data via cellxgene census. I stilll get sessionInfo |
|
@mihem thank you -- can you please show us the output of (Actually I can already see in the above that you have tiledb 0.21.1 whereas for tiledbsoma 1.4.3 we need to be at 0.20.3.) You should be able to do Then re-enter R. |
|
Phew. I am truly sorry for (as a card-carrying member of 'team Now, the segfaults. John, Paul H, and I have seen that too. The harder part is that the package tail gets longer. One issue we had recently is that package An old school alternative to the |
|
Ah, sorry, you already explained that earlier. So after downgrading to 0.20.3 it finally works 😂 . So no ccache and tiledb version 0.20.3. Not very intuitive, so maybe a note on the homepage for users would be useful (of course not if you solve it soon in the next version). (Hehe. Wouldn't never have thought that you were not right from the very beginning about an issue with C++ in R) Thanks for your great support, both of you! |
|
We are truly sorry for this. But we are working really really hard on getting the SOMA 1.5.0 release out, with feature parity between Python and R, and as it goes found a bug to fix here to there but it is ... "really close". Should be simpler afterwards. The core issue here is a somewhat direct attempt at version pinning but as you noticed there are different versions under the hood. It's tricky. When everything is aligned it ... just works (TM) as it should. |
|
No need to apologize, I absolutely understand. I just saw this post on Twitter And thought I give it a quick shot .. and other biologist/physicians with limited C++ knowledge might too. I really appreciate that you also support R because Seurat is still the most popular tool in the single cell community I think. |
Describe the bug
When running the R code provided on this page https://chanzuckerberg.github.io/cellxgene-census/cellxgene_census_docsite_quick_start.html#installation:
the following error occurs
Environment
Additional context
Not sure if this is related, but I had problem installing
tiledbsoma.failed with
So i used @eddelbuettel tip from here single-cell-data/TileDB-SOMA#1599
and downloaded the binary tiledbsoma_1.4.3-jammy.tar.gz from here https://chanzuckerberg.r-universe.dev/tiledbsoma which worked fine as did the cellxgene.census.
The text was updated successfully, but these errors were encountered: