Command line use (and testing) of taxsim.js file

[thomascwells]

Context: I'd like to use your compiled .wasm and .js files offline as part of an R package. [1]

Issue: I can't seem to figure out how to interact with the js or wasm you built via CLI.

I tried the test line from your make file, but that returned an error.

>echo "year,mstat\n2020,2" | node -e "require('./taxsim.js')()"
 TAXSIM: ERROR year,mstat\n2020 is not a taxsim35 input variable.
STOP 901

So I tried a different form of the input (this from the examples on NBER), but still no luck.

>echo "tax,mstat,year,ltcg,idtl\n1,2,1970,100000,5" | node -e "require('./taxsim.js')()"
 TAXSIM: ERROR tax,mstat,year,l is not a taxsim35 input variable.
STOP 901

If I switch to Object-style input, I get a different output (STOP 999), but not the results.

>echo "{taxsimid: 1, mstat: 2, year: 1970, ltcg: 100000, idtl: 5}" | node -e "require('./taxsim.js')()"
STOP 999

I've also tried using wasmer but I think it fails because of some of the emscripten compile settings. I don't think wasmer and other WebAssembly runtimes know how to use the .js.

$ wasmer taxsim.wasm
error: failed to run `taxsim.wasm`
╰─▶ 1: Emscripten requires at least one imported table

My ideal situation would be to use a JS or WASM interface already available in R, like the V8 port for R, but my attempts to make that work have hit dead ends as well.

> ctx <- V8::v8()
> ctx$source("taxsim.js")
> ctx$call('taxsim', '{taxsimid: 1, mstat: 2, year: 1970, ltcg: 100000, idtl: 5}')
Aborted(Assertion failed: shell environment detected but not enabled at build time.  Add 'shell' to `-s ENVIRONMENT` to enable.)
named list()

Do you have any suggestions?

I'm mostly an R programmer just trying to fuddle my way through this, so apologies if I've made some obvious beginner mistake.

[1] I'd like to add a feature to @shanejorr's package so calculations can be done without sending data to a remote server.

[tmm1]

Hi! Glad to see someone playing with this.

Try echo -e "..." | ...

node should work, as should v8. Wasmer likely will not.

---

[tmm1@tmm1-nuc ~]$ echo "\n\n"
\n\n
[tmm1@tmm1-nuc ~]$ echo -e "\n\n"

[tmm1]

Add 'shell' to -s ENVIRONMENT to enable.

This might be simple enough to add to the Dockerfile. I'll research the emscripten docs to understand the implications.

[thomascwells]

Try echo -e "..." | ...

Ah ha! Yes, that works! The -e does seem to be important.

But also, I was trying to run my commands on Windows powershell (facepalm), but the newline is different there.

Powershell:
echo "year,mstat`n2020,2" | node -e "require('./taxsim.js')()"

Bash:
echo "year,mstat\n2020,2" | node -e "require('./taxsim.js')()"

Thanks for that!

Enabling the shell environment would help things a lot. Another (additional? complementary?) option might be standalone WASM for use with wasmer, as discussed here: https://v8.dev/blog/emscripten-standalone-wasm. However, I don't know if that would break integration with your taxsim.app implementation (which is quite nice, by the way!).

[tmm1]

Thanks for the link.

If you want to run both on the Web and elsewhere, and you want 100% optimal code size and startup times, you should make two separate builds, one with -s STANDALONE and one without.

This isn't ideal but may be required for wasm runtimes.

Enabling the shell environment would help things a lot.

I poked at this and looks like we can pass a list into ENVIRONMENT:

https://github.com/emscripten-core/emscripten/blob/4881eaf6c23f590e5a8f8bd0d013c6c1cc11f42f/emcc.py#L288-L295

I'm not sure what the default is but it must be web,node since both of those work.

[tmm1]

I made a change in 5530026

The generated js file didn't really change much. You can find it at https://taxsim.app/taxsim.js. Let me know how it works out.

[tmm1]

I tried a quick STANDALONE_WASM=1 build and ran it with both wasmtime and wasmer. Unfortunately neither worked.

$ wasmer taxsim.wasm
error: failed to run `taxsim.wasm`
╰─▶ 1: Error while importing "wasi_snapshot_preview1"."args_sizes_get": unknown import. Expected Function(FunctionType { params: [I32, I32], results: [I32] })

$ wasmtime taxsim.wasm
Error: failed to run main module `taxsim.wasm`
Caused by:
    0: failed to instantiate "taxsim.wasm"
    1: unknown import: `env::__syscall_dup` has not been defined

My understanding is that WASI is still in early days so I've been waiting for the standard to finalize before digging in further.

There is some investigation happening here as well: StarGate01/Full-Stack-Fortran#7

Also we're behind a few versions of emscripten, which is blocked on investigation of this issue: StarGate01/Full-Stack-Fortran#10 (comment)

[tmm1]

I did some testing with d8

Before 5530026, I see the same issue as you:

$ d8 -e "load('./taxsim.js'); taxsim()"
Aborted(Assertion failed: shell environment detected but not enabled at build time.  Add 'shell' to `-s ENVIRONMENT` to enable.)

After the change, it gets past that:

$ printf "year,mstat\n2020,2" | d8 -e "load('./taxsim.js'); taxsim()"
STOP 993

Somehow the input requirements are different.. it must not be receiving EOF correctly? If I add an extra newline, it returns the data atleast before erroring:

$ printf "year,mstat\n2020,2\n" | d8 -e "load('./taxsim.js'); taxsim()"

taxsimid,year,state,fiitax,siitax,fica,frate,srate,ficar,tfica
0.,2020,0,-3600.00,0.00,0.00,-7.65,0.00,15.30,0.00
STOP 993

[tmm1]

I made some more progress in cdd1355

However I think it will take a bit more finagling to get it working inside R/v8.

https://github.com/emscripten-core/emscripten/blob/e1d97f55d12ed63dc30e547c818af9a63be0b976/src/shell.js#L317-L327

[tmm1]

It works in R!

> download.file("https://taxsim.app/taxsim.wasm", "taxsim.wasm")
> wasm <- readBin("taxsim.wasm", raw(), file.info("taxsim.wasm")$size)

> library(V8)
Using V8 engine 9.6.180.12
> ctx <- v8()
> ctx$assign("wasmBinary", wasm)
> ctx$source("https://taxsim.app/taxsim.js")

> ctx$call("taxsim", JS("{year:2020, mstat:2}"), JS("{wasmBinary}"), await=TRUE)
[1] "taxsimid,year,state,fiitax,siitax,fica,frate,srate,ficar,tfica\r\n0.,2020,0,-3600.00,0.00,0.00,-7.65,0.00,15.30,0.00\r\n"

[feenberg]

This is very nice. We should make it easier to get a copy of taxsim. I am probably going to have to move the taxsim ftp interface off the NBER server because our security auditor is uneasy with it. The new location will be at taxsim.us, and I will make all of taxsim available there as soon as I have it set up, including taxsim.wasm. I suppose if someone is running taxsim.wasm in the browser, they can be confident taxsim.wasm doesn't do anything bad to there system. Does node protect them? nber.org had the advantage of being clearly legitimate (at least to the typical potential user). Someone might be suspicious of a new host. A lot of my users are in the US government, where security would be a particular issue. Thanks so much! Daniel Feenberg 617-682-6204

…

On Thu, 9 Jun 2022, Thomas Wells wrote: Try echo -e "..." | ... Ah ha! Yes, that works! The -e does seem to be important. But also, I was trying to run my commands on Windows powershell (facepalm), but the newline is different there. Powershell: echo "year,mstat`n2020,2" | node -e "require('./taxsim.js')()" Bash: echo "year,mstat\n2020,2" | node -e "require('./taxsim.js')()" Thanks for that! Enabling the shell environment would help things a lot. Another (additional? complementary?) option might be standalone WASM for use with wasmer, as discussed here: https://v8.dev/blog/emscripten-standalone-wasm. However, I don't know if that would break integration with your taxsim.app implementation (which is quite nice, by the way!). — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to thisthread.[AB55AVNNZANWIWZELCVRST3VOKPCVA5CNFSM5YLYG4Z2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJK TDN5WW2ZLOORPWSZGOISTZTPI.gif] Message ID: ***@***.***>

[feenberg]

On Thu, 9 Jun 2022, Aman Gupta Karmani wrote: Thanks for the link. If you want to run both on the Web and elsewhere, and you want 100% optimal code size and startup times, you should make two separate builds, one with -s STANDALONE and one without.

Where do I put that? Is the savings significant? If it is, I want to offer both. If it is minor, perhaps we can avoid the complication. I just did a test. .37 seconds to load and process a single record. .39 to process 1,500 records. .37 is pretty good, by modern standards, I think.

This isn't ideal but may be required for wasm runtimes. Enabling the shell environment would help things a lot. I poked at this and looks like we can pass a list into ENVIRONMENT: https://github.com/emscripten-core/emscripten/blob/4881eaf6c23f590e5a8f8bd0d013c6c1cc11f42f/em cc.py#L288-L295

taxsim doesn't use any environment variables that I am aware of. What makes this useful? I just noticed that a blank line in the data can be a problem. I'll fix that too. Daniel Feenberg NBER

…

I'm not sure what the default is but it must be web,node since both of those work. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to thisthread.[AB55AVPIHJZEPVG3EQZHB53VOKR3PA5CNFSM5YLYG4Z2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJK TDN5WW2ZLOORPWSZGOIST5ZVI.gif] Message ID: ***@***.***>

[feenberg]

On Thu, 9 Jun 2022, Aman Gupta Karmani wrote: I did some testing with d8 Before 5530026, I see the same issue as you: $ d8 -e "load('./taxsim.js'); taxsim()" Aborted(Assertion failed: shell environment detected but not enabled at build time. Add 'shell ' to `-s ENVIRONMENT` to enable.) After the change, it gets past that: $ printf "year,mstat\n2020,2" | d8 -e "load('./taxsim.js'); taxsim()" STOP 993 Somehow the input requirements are different.. it must not be receiving EOF correctly? If I add an extra newline, it returns the data atleast before erroring: $ printf "year,mstat\n2020,2\n" | d8 -e "load('./taxsim.js'); taxsim()" taxsimid,year,state,fiitax,siitax,fica,frate,srate,ficar,tfica 0.,2020,0,-3600.00,0.00,0.00,-7.65,0.00,15.30,0.00 STOP 993

I added a test to skip blank lines. New Fortran source is temporarily available from: http://www.nber.org/~feenberg/taxsim.f Dan

…

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to thisthread.[AB55AVOADDZVD7QZJIHAAR3VOLNBLA5CNFSM5YLYG4Z2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJK TDN5WW2ZLOORPWSZGOISVBOTI.gif] Message ID: ***@***.***>

[tmm1]

Thanks Dan!

The changes discussed above have all been incorporated into this repository. To use them, you can:

$ cd taxsim.js
$ git pull
$ cp /path/to/latest/taxsim.f .
$ make

This will create a new taxsim.{js,wasm} which is compatible with node and libv8, and can be used in R.

---

taxsim doesn't use any environment variables that I am aware of. What makes this useful?

This is a flag to the compiler which tells it to make the wasm/js compatible with libv8. It was incorporated in 5530026 so there's nothing you need to do.

two separate builds, one with -s STANDALONE and one without.
Where do I put that? Is the savings significant?

I tried this and it didn't work so there's no reason or benefit to this at the moment.

The new location
will be at taxsim.us, and I will make all of taxsim available there as
soon as I have it set up, including taxsim.wasm.

Thats great news!

If taxsim.f source is going to become public, we should host a copy here on GitHub as well, so it's easier to see how it changes over time. I'm happy to help with this once you get to that point.

I suppose if someone is running taxsim.wasm in the browser, they can be
confident taxsim.wasm doesn't do anything bad to there system. Does node
protect them?

Yes wasm in the browser is extremely sandboxed and very safe. It cannot access the filesystem, environment or really anything on your system.

node is less safe because it has no sandbox. Technically if the js was crafted maliciously it could access anything on your system, since node.js is a full programming language and runtime.

For users concerned about security, having the taxsim.f source available alongside the source in this repo would allow them to create their own copy of taxsim.wasm from scratch and audit every step of the process.

[tmm1]

@thomascwells Let me know if the R example works in your environment. I only tested on my macOS machine.

If you end up working on adding support to https://github.com/shanejorr/usincometaxes, please cc me on any pull requests. I'd also like to add an entry to the README here once that happens, under projects using taxsim.js.

[thomascwells]

It works in R!

This looks great! Thanks for all the updates Aman! I'll give this a test run today.

Somehow the input requirements are different.. it must not be receiving EOF correctly? If I add an extra newline, it returns the data at least before erroring:

When I was trying to use the HTTP interface via curl as described here*, it would fail if the uploaded file contained an end-of-file new line. I had to create a file without a final blank line or I would get errors each time. Some working R code here, but Dan's recent change for skipping blank lines might solve that issue.

*Note: I think the URL in the curl example on NBER didn't work for me. It should be https://taxsim.nber.org/uptest/webfile.cgi, not http://wwwdev...

---

I was part way through crafting some responses to Dan's questions and comments, but Aman beat me to it with better and more thorough answers. Suffice to say, I agree with Aman's thoughts about sandboxing and the importance of being able to audit the compilation process from source. I'm also happy to help with GitHub publishing & maintenance if you go that way. Building some CI/CD workflows triggered on updates to the Fortran source would be ideal!

Dan, I know you had some concerns about possible derivative work if you were to broadly publish the source. Do you think those concerns could be addressed with appropriate licensing?

[thomascwells]

Unfortunately, I could not reproduce your success in R. This is what I got:

> download.file("https://taxsim.app/taxsim.wasm", "taxsim.wasm")
trying URL 'https://taxsim.app/taxsim.wasm'
Content type 'application/wasm' length unknown
downloaded 900 KB

> wasm <- readBin("taxsim.wasm", raw(), file.info("taxsim.wasm")$size)
> library(V8)
> ctx <- v8()
> ctx$assign("wasmBinary", wasm)
> ctx$source("https://taxsim.app/taxsim.js")
> ctx$call("taxsim", JS("{year:2020, mstat:2}"), JS("{wasmBinary}"), await=TRUE)
failed to asynchronously prepare wasm: CompileError: WasmCompile: Wasm decoding failed: unknown section code #0x10 @+1443
Aborted(CompileError: WasmCompile: Wasm decoding failed: unknown section code #0x10 @+1443)
Error in context_eval(join(src), private$context, serialize, await) : 
  RuntimeError: Aborted(CompileError: WasmCompile: Wasm decoding failed: unknown section code #0x10 @+1443). Build with -s ASSERTIONS=1 for more info.

I can continue trying some things, including rebuilding the WASM using the emscripten stack with the copy of the fortran source (Thanks Dan!)

However, it's worth noting that base-R has capabilities for compiling Fortran and running those modules directly, so maybe we're trying to solve the wrong problem.

• https://stat.ethz.ch/R-manual/R-devel/library/base/html/Foreign.html
• https://bookdown.org/yihui/rmarkdown/language-engines.html#c-and-fortran

Should we fork the R-related components of this conversation to a different issue? Or possibly over to the R package (usincometaxes) repo?

[tmm1]

CompileError: WasmCompile: Wasm decoding failed: unknown section code #0x10 @+1443

My hunch is this might be related to the version of libv8? What did library(V8) output? Mine was:

library(V8)
Using V8 engine 9.6.180.12

Actually, you said you're on windows so there might be some line encoding issue going on. See emscripten-core/emscripten#7987 (comment)

There may be some other variant of download.file that ensures binary mode?

On https://stat.ethz.ch/R-manual/R-devel/library/utils/html/download.file.html it says:

The choice of binary transfer (mode = "wb" or "ab") is important on Windows, since unlike Unix-alikes it does distinguish between text and binary files and for text transfers changes \n line endings to \r\n (aka ‘CRLF’).

So I think the fix is simply:

> download.file("https://taxsim.app/taxsim.wasm", "taxsim.wasm", mode="wb")

[tmm1]

However, it's worth noting that base-R has capabilities for compiling Fortran and running those modules directly, so maybe we're trying to solve the wrong problem.

I believe Dan has also done some similar work in the past, to build a stata/ado plugin using the fortran code directly.

You're welcome to pursue this, but I imagine it would require changes to the fortran code itself to make it behave correctly in a plugin environment. Then you would also need to compile the plugin for different operating systems separately.

Personally I prefer the wasm approach, since it can treat the fortran source as a complete black box and results in a cross-platform artifact.

[thomascwells]

Yes, it appears we are using different versions of libv8:

> library(V8)
Using V8 engine 6.2.414.50

... but, the download.file( ... mode="wb" fix worked like a charm!

> download.file("https://taxsim.app/taxsim.wasm", "taxsim.wasm", mode="wb")
trying URL 'https://taxsim.app/taxsim.wasm'
Content type 'application/wasm' length unknown
downloaded 900 KB

> wasm <- readBin("taxsim.wasm", raw(), file.info("taxsim.wasm")$size)
> library(V8)
Using V8 engine 6.2.414.50
> ctx <- v8()
> ctx$assign("wasmBinary", wasm)
> ctx$source("https://taxsim.app/taxsim.js")
> ctx$call("taxsim", JS("{year:2020, mstat:2}"), JS("{wasmBinary}"), await=TRUE)
[1] "taxsimid,year,state,fiitax,siitax,fica,frate,srate,ficar,tfica\r\n0.,2020,0,-3600.00,0.00,0.00,-7.65,0.00,15.30,0.00\r\n"

I'll do some more testing with this. If it works well, I'll try to get this wasm+v8 method added to the R package.

[feenberg]

On Fri, 10 Jun 2022, Aman Gupta Karmani wrote: However, it's worth noting that base-R has capabilities for compiling Fortran and running those modules directly, so maybe we're trying to solve the wrong problem. I believe Dan has also done some similar work in the past, to build a stata/ado plugin using the fortran code directly.

I did make a stata plugin for taxsim9. The plugin avoids a translation from binary to ascii, but I can think of no other advantage. I was able to make it work, but found debugging extremely difficult, as errors from the fortran were never displayed to me, nor could I write code that displayed error messages from Fortran. In the end I didn't want to continue on such a fraught path. I can send you what I wrote but I think wasm is a superior approach. R may be different, but what are you thinking of for the user interface? Would it be a function that took a 35xN array and returned a 46xN array? It wouldn't be difficult to extract some of the code in taxsim.f to create a fortran function that did that. Or would it take and return a single taxpayer? Even easier. Could R handle taking a subset of input variables and passing the complete set to the Fortran code? If you are really eager to go this way, we should have a talk on the telphone. A conversation is really the only way to handle something like this. Daniel Feenberg ***@***.***

…

You're welcome to pursue this, but I imagine it would require changes to the fortran code itself to make it behave correctly in a plugin environment. Then you would also need to compile the plugin for different operating systems separately. Personally I prefer the wasm approach, since it can treat the fortran source as a complete black box and results in a cross-platform artifact. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because youcommented.[AB55AVL2SDEHXEVWCGQMBZTVONZI7A5CNFSM5YLYG4Z2YY3PNVWWK3TUL52HS4DF VREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOISZNCQI.gif] Message ID: ***@***.***>

[thomascwells]

I'll give you a call later today about the current state of and my goals for the R functions!

[feenberg]

I don't understand git or docker, but I am confused by the instructions below. Do I need to cd .git git stash before running git pull? I also tried "git commit" but got an error. I have forgotten what I changed in Dockerfile that caused a problem. Is there a way to get the original Dockerfile? Also, when I run git pull, it downloads 400 megabytes of stuff. What is all that? I thought it only needed to fetch taxsim.js. Even after this, taxsim.js only ends up with 15 bytes and taxsim.wasm isn't changed. Please give me a bit of help here. Possibly a telephone call would be appropriate. My cellphone is below. I'll be in the office till 5pm in Boston, and can take a call any other time too. Daniel Feenberg 617-682-6204

…

On Fri, 10 Jun 2022, Aman Gupta Karmani wrote: Thanks Dan! The changes discussed above have all been incorporated into this repository. To use them, you can: $ cd taxsim.js $ git pull $ cp /path/to/latest/taxsim.f . $ make This will create a new taxsim.{js,wasm} which is compatible with node and libv8, and can be used in R. ____________________________________________________________________________ taxsim doesn't use any environment variables that I am aware of. What makes this useful? This is a flag to the compiler which tells it to make the wasm/js compatible with libv8. It was incorporated in 5530026 so there's nothing you need to do. two separate builds, one with -s STANDALONE and one without. Where do I put that? Is the savings significant? I tried this and it didn't work so there's no reason or benefit to this at the moment. The new location will be at taxsim.us, and I will make all of taxsim available there as soon as I have it set up, including taxsim.wasm. Thats great news! If taxsim.f source is going to become public, we should host a copy here on GitHub as well so it's easier to see how it changes over time. I'm happy to help with this once you get to that point. I suppose if someone is running taxsim.wasm in the browser, they can be confident taxsim.wasm doesn't do anything bad to there system. Does node protect them? Yes wasm in the browser is extremely sandboxed and very safe. It cannot access the filesystem, environment or really anything on your system. node is less safe because it has no sandbox. Technically if the js was crafted maliciously it could access anything on your system, since node.js is a full programming language and runtime. For users concerned about security, having the taxsim.f source available alongside the source in this repo would allow them to create their own copy of taxsim.wasm from scratch and audit every step of the process. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because youcommented.[AB55AVP5ZE2KGRTWO4JA6PDVONMGFA5CNFSM5YLYG4Z2YY3PNVWWK3TUL52HS4DF VREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOISYVPEA.gif] Message ID: ***@***.***>

[thomascwells]

I'll give you a call right now. We can discuss the git commands and R.

[thomascwells]

Just a note in case someone else comes across this thread: most of the discussion related to command line use of TAXSIM and interfacing with R has shifted to an internal NBER Gitlab repo. I expect whatever comes of that discussion will eventually work its way back to this repo, but for now I guess this "issue" is just on-hold.

[feenberg]

On Thu, 9 Jun 2022, Thomas Wells wrote: Context: I'd like to use your compiled .wasm and .js files offline as part of an R package. [1] Issue: I can't seem to figure out how to interact with the js or wasm you built via CLI. I tried the test line from your make file, but that returned an error. >echo "year,mstat\n2020,2" | node -e "require('./taxsim.js')()" TAXSIM: ERROR year,mstat\n2020 is not a taxsim35 input variable. STOP 901

Clearly \n is not being translated into a newline. You could try cat'ing a two line file and see if that works any better. Clearly taxsim is reading the input line as a character string, but is also having trouble dividing it into words at each comma. The fortran read statement is: read(a,*,end=7,iostat=iostat) dnames where variable "a" holds the line of text and dnames is a one dimensional character array. The error message is written by: write(*,*) "TAXSIM: ERROR ",dnames(i), & " is not a taxsim35 input variable." which should print only one element of dnames. I have no thoughts about why the read failed to separate the variable names into separate elements of dname. Daniel Feenberg 617-682-6204

…

So I tried a different form of the input (this from the examples on NBER), but still no luck. >echo "tax,mstat,year,ltcg,idtl\n1,2,1970,100000,5" | node -e "require('./taxsim.js')()" TAXSIM: ERROR tax,mstat,year,l is not a taxsim35 input variable. STOP 901 If I switch to Object-style input, I get a different output (STOP 999), but not the results. >echo "{taxsimid: 1, mstat: 2, year: 1970, ltcg: 100000, idtl: 5}" | node -e "require('./taxsim .js')()" STOP 999 I've also tried using wasmer but I think it fails because of some of the emscripten compile settings. I don't think wasmer and other WebAssembly runtimes know how to use the .js. $ wasmer taxsim.wasm error: failed to run `taxsim.wasm` ╰─▶ 1: Emscripten requires at least one imported table My ideal situation would be to use a JS or WASM interface already available in R, like the V8 port for R, but my attempts to make that work have hit dead ends as well. > ctx <- V8::v8() > ctx$source("taxsim.js") > ctx$call('taxsim', '{taxsimid: 1, mstat: 2, year: 1970, ltcg: 100000, idtl: 5}') Aborted(Assertion failed: shell environment detected but not enabled at build time. Add 'shell ' to `-s ENVIRONMENT` to enable.) named list() Do you have any suggestions? I'm mostly an R programmer just trying to fuddle my way through this, so apologies if I've made some obvious beginner mistake. [1] I'd like to add a feature to @shanejorr's package so calculations can be done without sending data to a remote server. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to thisthread.[AB55AVOYULQGK6ZDKGPODQDVOKBGDA5CNFSM5YLYG4Z2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45 C7NFSM4S4BVPGQ.gif] Message ID: ***@***.***>