dorianps
commented
4 years ago Hi Lars, I will need to look better into this by next week. From your description looks the error comes when a complete Linda folder exists for the subject, but works in all other cases. If that is the case, I can try to reproduce the error and check the best workaround. But let me know if my understanding is correct at the mean time.
Dorian
On Fri, Aug 30, 2019, 3:19 PM Lars notifications@github.com wrote:
Dear Dorian,
I am currently building a code wrapper for LINDA (v0.5.0) for a project and I encountered an "issue" when using the "cache = TRUE" option in "linda_predict".
The idea behind the wrapper is that I have multiple stroke cases arranged in BIDS format which I want to run (and re-run parts if something went wrong) continuously without stopping. The problem is that when the "cache = TRUE" option it will stop the whole sequence of code if the segmentation of one case is already finished, which is not ideal for a project wrapper.
The error message returned by R states that it "could not find object L", which I figured was related to the last "return(L)" statement in "linda_predict". This is what I suspect is the core issue. Because after I commented it out in my own "linda_predict_custom", sourced from the package::function LINDA::linda_predict, it then ran for a while until additional errors occurred.
I dont know the extent of work required to compile and change R packages, I will therefore also include a possible workaround. It is a workaround that I have tested to work.
I worked on a possible workaround with a custom "linda_predict" function outside of the LINDA package, but that has proven to be a less than ideal approach. Although by only commenting out "return(L)" at the end of the function, additional (new) errors followed, such as:
- Not finding the "print_msg" function. Which I only found in "dorianps/LINDA/R/aaa_utils.R",
- And not finding functions found in ANTsRCore, and possibly also in ANTsR and ITKR too.
These issues were simple enough to correct, although not elegantly, to make "linda_predict_custom" work:
- Copy-pasting the "print_msg" function into Rstudio environment (or in the same file as the R script which is sourced together with "linda_predict_custom").
- Explicitly requiring and loading the libraries ANTsRCore alongside ANTsR and ITKR before running "linda_predict_custom".
I hope this ratched implementation can be helpful to others that also want to build a wrapper for LINDA for whatever reason.
Do note: As "linda_predict" writes directly to the system I did not personally have need for the output returned by the function. This is and important element for this workaround to be worth trying.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/dorianps/LINDA/issues/24?email_source=notifications&email_token=ACFJU7J55RIS26ZXFFHCG3TQHEM4XA5CNFSM4ISNHZ22YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HIOLHVA, or mute the thread https://github.com/notifications/unsubscribe-auth/ACFJU7MJHFZKZ2CN4ORTPIDQHEM4XANCNFSM4ISNHZ2Q .
labhstats
Dear Dorian,
I am currently building a code wrapper for LINDA (v0.5.0) for a project and I encountered an "issue" when using the "cache = TRUE" option in "linda_predict".
The idea behind the wrapper is that I have multiple stroke cases arranged in BIDS format which I want to run (and re-run parts if something went wrong) continuously without stopping. The problem is that when the "cache = TRUE" option it will stop the whole sequence of code if the segmentation of one case is already finished, which is not ideal for a project wrapper.
The error message returned by R states that it "could not find object L", which I figured was related to the last "return(L)" statement in "linda_predict". This is what I suspect is the core issue. Because after I commented it out in my own "linda_predict_custom", sourced from the package::function LINDA::linda_predict, it then ran for a while until additional errors occurred.
I dont know the extent of work required to compile and change R packages, I will therefore also include a possible workaround. It is a workaround that I have tested to work.
I worked on a possible workaround with a custom "linda_predict" function outside of the LINDA package, but that has proven to be a less than ideal approach. Although by only commenting out "return(L)" at the end of the function, additional (new) errors followed, such as:
These issues were simple enough to correct, although not elegantly, to make "linda_predict_custom" work:
I hope this ratched implementation can be helpful to others that also want to build a wrapper for LINDA for whatever reason.
Do note: As "linda_predict" writes directly to the system I did not personally have need for the output returned by the function. This is and important element for this workaround to be worth trying.