Currently errors are serialized into a single text line with some structure, but no recognizable format, like this:

Error: ShellError { error: Diagnostic(ShellDiagnostic { diagnostic: Diagnostic { severity: Error, code: None, message: "Cannot canonicalize file "/file.json" because Custom { kind: Other, error: "operation not supported on wasm yet" }", labels: [Label { style: Primary, file_id: 0, range: 5..14, message: "Cannot canonicalize" }], notes: [] } }), cause: None }

In this case for running open file.json.

Jonathan mentioned the option to serialize errors as JSON, so that they can be formatted on the JS side.

Hi there :)

How complicated would it be, to update this to a current version?

Is there any chance? 🙏🏻

Looks like there are some tips here: https://rustwasm.github.io/book/reference/code-size.html

For how we might be able to optimize the code size of the wasm we transmit.

on the cli:

random dice -d 10 -s 12


───┬────
 0 │  4
 1 │  2
 2 │ 12
 3 │  4
 4 │  1
 5 │  2
 6 │  5
 7 │ 12
 8 │  1
 9 │  8
───┴────

here:

7341010121261212

Looks like this applies to all lists, e.g.

echo "[1,2,3]" | from json

results in 123

Running random bool panicks and only dumps that error with console.error

panicked at 'could not initialize thread_rng: getrandom: this target is not supported', /Users/jzaefferer/.cargo/registry/src/github.com-1ecc6299db9ec823/rand-0.7.3/src/rngs/thread.rs:65:17

I guess the console.error comes from the console_error_panic_hook crate

Is there some way to have the wasm.run_nu Promise rejected instead, to handle the error in our "application" code?

If not, can we deal with panics on the Rust side and forward them through the existing serialized error channel?

It also looks like something "exits" once this type of error occurs, and running other commands isn't possible anymore (a page reload helps).

Maybe using the widespread octocat banner in the top right

This is what I'm thinking. The UI something like this now.
[button1] [button2] [button3]
text box command
results

If it worked more like a notebook ala jupyter it would have
text box command
results

text box command
results

text box command
results

And so on. You could scroll up and down through all your text box commands & results and execute them, change them or whatever. Or, if we were really slick, the output of one result could be stored as a variable and used in the next text box command just like jupyter.

It would be nice to have this on the demo site and the main site - found on http://cheats.rs (see bottom right).

Source code here https://github.com/ralfbiedert/cheats.rs

Some ideas for exploring tables in the browser:

For table headers:

• select column => Appends | get {name} to the current command and reruns it.
• group-by column, append | group-by {name}
• sort-by column, append | sort-by {name}

A pivot shortcut already shows up when the table is too wide.

..to be continued...

Until this is properly integrated into the nushell website, we should deploy it on GitHub Pages, to have a public preview.

Probably use GitHub Actions to build from mastmain.

I seem to get it into a state where no command is interpreted any longer. I can click the top buttons and they put text in the window but if I hit run or ctrl/cmd-enter nothing happens. There are no errors visible. So, would it be possible to have some type of a reset button? I'm not sure what that would do since when this happens today I have to ctrl-c out and do npm start again.

At the time of writing, the demo seems to be using Nushell v0.38, so lots of the more recent features won't be available. It would be nice if the version (and possibly the commit date as well) was indicated somewhere in the page, e.g. in the footer.

We may want to play with the size of columns a bit. Just did help commands and got this (notice the command name is hard to read):

I was running through the demo at https://www.nushell.sh/demo/ and was contacted by the infosec team at my firm regarding "nuvfs.zip" and was informed that it matched multiple malicious signatures.

The infosec team requested I no longer continue accessing the website and refrain from investigating nushell. I did request permission to file an issue regarding the matter, as I doubt this was intentional. Either the wasm module has actually been infected or there is some pattern employed in the layout of the VFS which is triggering a security rule.

Unfortunately, I have no additional details I am able to provide.

It would be pretty helpful for us to see what errors demo users run into, to prioritize fixes.

I have plenty experience with Sentry, so I'd check their free/OSS options first.

I'd make sure to enable all anonymization options - we don't need to collect anyone's IP address. Only basic user agent info to reproduce issues.

When loading examples, it would be nice to be able to go back to the previous input. Two undo/redo buttons that keep track of the stack and allow going back and forth would work.

https://github.com/ArthurClemens/Javascript-Undo-Manager looks fine for that.

Maybe we could have a config.toml in a special BrowserFS folder like /nu/config/config.toml and then load that config so we have some sort of environment variables and aliases to play around with.

Can you bring over the new to html code that was committed yesterday? I'd like to see what the new to html --html_color and to html --no_color does for things like this.

Having a playground for opening, selecting and modifying files in the browser makes this project pretty interesting, but right now there's no files, and open generally doesn't work in wasm.

There might be some way to mock a few files (json, csv, toml), at least enough for ls and open to do something useful.

We use auto-rotation in the terminal to help fit things on the screen when it's a single long row. Feels like we might want to consider that for wasm-nu also:

Notice size of scrollbar.

To make this more than a slightly crippled demo for nushell, we could offer a textarea (and upload/dnd section) for providing JSON, toml, csv or other data that nushell can parse (maybe auto detect the format), then a pipeline for processing it and an output area for the result, as HTML table and whatever structured output is desired (default to input format, can switch to something else.

The format selects could be fed with a list of subcommands from "from" and "to".

Since everything runs on the client, we could update everything automatically while typing, as long as the pipeline is a valid command

Was thinking that maybe there should be a readme button somewhere on the page that tells what's working and what's not working (i.e. @jzaefferer 's issue that lists it). Error handling is still in it's infancy and I don't want people trying it and getting aggravated and then assuming nushell as a whole doesn't support all the commands we claim it does.

How about a more complicated example blue button to show a little more power?

open samples/movies.csv | group-by Year | get "2010" | update "Worldwide Gross" { get 'Worldwide Gross' | str trim --char '$' | str to-decimal } | sort-by 'Worldwide Gross' | reverse | update 'Worldwide Gross' { get 'Worldwide Gross' | str from -d 2 | format "${$it}" }

I wish we could write it with comments like:

open samples/movies.csv     # open the file as a csv
| group-by Year             # group the data by Year
| get "2010"                # get only the column labeled "2010"
| update "Worldwide Gross"  # update the Worldwide Gross column
{ get 'Worldwide Gross'     # get the Worldwide Gross column
| str trim --char '$'       # remove the $ character
| str to-decimal }          # convert the string numbers to actual decimal numbers
| sort-by 'Worldwide Gross' # sort by these newly created decimal numbers in the Worldwide Gross column
| reverse                   # reverse the order putting the largest numbers at the top
| update 'Worldwide Gross'  # update/overwrite the Worldwide Gross column again
{ get 'Worldwide Gross'     # get the data from the Worldwide Gross column
| str from -d 2             # convert the decimals to strings with 2 decimal places
| format "${$it}" }         # expand $it with the format command to add the $ back again

For example, each --help shows three examples. I want to click on those (or an extra button next to them) to execute them.

alias: alias say-hi [] { echo 'Hello!' }; say-hi => error: Failed to spawn process

autoenv: might be possible to emulate with #3 or #34, but seems like overkill

cal, date: panicked at 'not implemented', ~/.cargo/registry/src/github.com-1ecc6299db9ec823/time-0.1.43/src/sys.rs:101:9

cd, cp, du, enter, exit, mkdir, mv, pwd, rm, save, touch: depends on #34

clear is meaningless in the browser

config: not supported in wasm; not sure how useful this really is though

debug: not sure how this works, since all the examples are using FS. Simply running debug outputs �[38;5;10mnothing.

describe: similar to debug, it partially outputs ansi color codes, e.g. for echo '[{"name":"Bob","age":25},{"name":"Fred","age":35}]' | from json | describe it outputs this:

empty?: not sure how to use this, since neither inline docs nor the website have examples (no docs for the command there at all)

fetch: 'failed to spawn'. We could make a custom implementation using the browser's fetch API.

history: error: Could not open history - maybe we can create that file with the vfs and update it with each command that the user runs?

kill: not supported in wasm; could emulate processes as web workers, list them with ps and kill them.

lines: the example shells out to native echo (^echo "two\nlines" | lines), which doesn't work here. I'm not sure how else to use lines

n/p: next, previous shell, doesn't do anything, not even output an error, though that also applies for the CLI. Help text doesn't explain how to use these.

random bool: needs a custom implementation like random dice

run_external: this looks like it makes no sense in this context, and should show a proper error message. Currently run_external echo gives back Failed to spawn process.

shells: This actually works, but without other shell-related commands, its difficult to test properly.

shuffle: fails badly (#12), probably same issue as with random bool.

split-by: no example, no markdown docs, and based on the description I have no idea how to use this

table: I'm not sure what this actually does, since, for example, sys and sys | table seem to do the same thing. Though here it breaks badly (#12)

with-env: error: operation not supported on wasm yet

Loading the several MB of wasm and the zip file can take a while, and until then the demo shows up, but isn't usable. A simple loading indicator would help

Maybe via https://codepen.io/dsenneff/pen/jOWdvqx

#3 provided a basic virtual file system, with support for ls and open.

ls only provides names and no other details, which is a bit of a problem since so many of our examples use various columns from ls.

Other fs related commands need to be replaced to use the vfs: cd, cp, du, enter, exit, mkdir, mv, pwd, rm, save, touch (maybe more)

ls *.foo doesn't work in wasm. Doesn't look like we support any wildcards (yet?)

Nu makes syntax highlighting available internally, so in theory we could expose this to wasm-nu, and then use the syntax highlighting information to color the user input.

So long as we don't interfere with them typing, it should be okay?

Now that the alternating rows are a bit lighter, it's harder to see the difference between them. I think having at least some alternative color would be nice to make the tables more readable.

Here's what it looks like for me in Linux/Firefox