Not just long. For some reason C has strtol() and not strtoi()?

And atoi() is broken.

These are wrappers around strtoll() and strtol()

Although we should still test for truncation f...

- Test it in mycpp/examples
- Change StringToInteger helper to take int64_t, and change all usages

We don't require arbitrary precision right now, since no shell has that.

bash, dash, and zsh al...

Fix lint error

We don't need it! It caused a lot of duplication.

Need to rewrite the evaluators to use them.

And maybe reduce the mylib.{py,pyi} duplication.

Add examples/integers.py

This seems like it can work without too much effort.

[benchmarks] Polish CPython configure diff

It seems like the (cached) differences are gone now?...

Fixes a case in spec/redirect under osh-cpp.

This is because you can't throw an exception from ...

We can compute _pipeline_status -> status AFTER waiting for process
subs.

They are independent.

Assert invariant for pipelines.

Use the status variable.

Working on "can't throw from destructor" bug

In C++, it still doesn't work, i.e. with clean Oils 0.20.0 tarball:

~/src/oils/oils-for-unix-0....

For fixing the "can't throw exceptions in destructors" issue.

Compared to Python and awk. This could be a blog post.

On just 92K lines of code, shells are >...

Tell the user to use 'read -r' instead.

It has a design problem - unbuffered I/O on file descri...

Explain why in a couple places.

Still need to remove it from the code.

Happens in Python too

There don't appear to be any hard errors for overflow in JS and Python.
Python has arbitrarily b...

e.g. I realized bash [[ a < b ]] should use libc strcoll()

Still need a solution for case foldi...

Most of it was done a long time ago.

This fixes the 'pp proc' bug in C++.

doc_token would be initialized as a separate var inside th...

They use common routines in core/num.py - integer divide and remainder.

This fixes a spec-cpp d...

No modulus like Python.

This is part of #1342.

This is part of #1342.

[demo] Survey division and modulus behavior. Next we'll make it
consistent between Python and C++.

Move assertion back to cpp/data_lang.cc.

Regenerate prebuilt/ files.

Factor out common ObjectId() function for ASDL types.

This is #1342. These cases complement the test cases for YSH modulus.

This avoids a compile warning. Those assertions should never happen.

Move another test case to spec/testdata. Avoid code on stdin.

Remove unnecessary pyos function.

Hm one is still failing?

Anything with read --line conflicts with the spec test framework, which
reads code on stdin.

j8...

Make 2 spec tests pass.

- introduce mylib::File and mylib::CFile
- CFile instead of CFileLine...

Move it to frontend/flag_util.py.

This fixes a couple bugs in test/spec-cpp.sh runfile builtin-...

No change in functionality.

Things that stand out:

- We should define Bool and Int as builtin ...

It works on hello.yaks!

Add to line counts

We still need the main wrapper

Document catch-22 with imports - def_write() vs. decl_write().

self.always_write()
self.def_write() # write in definitions only

I think we will add o...

It should be translatable now.

Also refactor to MustBeList().

Also:

* [ysh/builtins] ctx, error.FatalRuntime(3, ...) -> error.Expr(...)

Still have to do the param list.

The infix rule is still working pretty well. We can use it fo...

Add gen_cpp.py stub

The infix rule is working out nicely for signatures:

(func main [argv [List Str]] => Int
...

This duplicates what's in soil/worker.sh a bit. But it should make the
build more correct.

* change j"..." example to u'...'
* add example with ++
* Polish answers some more

--------...

And refine the yaks.asdl schema.

I think we have a strategy. Though it has several steps, and might take
quite awhile.

OK that wasn't too hard. Let's see if everything passes.

- More NIL8 parsing tests
- binary size metric: yaks is about 200K
- autoformat the code
- enabl...

(k:v k2:v2) is equivalent to
((: k v) (: k2 v2))

These infix operators should be useful...

Translates to C++.

Next: do special rule for infix operators?

Remove inverted dependencies from data_lang/j8.py.

There are some issues with using NIL8 as the syntax for both Yaks
and ASDL pretty printing. We ...

- Start nil8.asdl
- Import j8, and move vm.Value* functions to j8
- core.vm pulls in Oils, whi...

JSON/JSON8 and NIL8 now both reject trailing data.

[spec/ysh-json] Make note of bug with trailing data.

We shouldn't accept it. Added this to the...

Notes on NIL8 grammar

Document J8 comments. Shell # vs. JavaScript // is an annoying issue.

But I just found that //...

But not JSON

Reformat Python code.

- Narrow Intermediate Language
- NIL8 Isn't Lisp

* ctx push, set, emit
* mini-argparse (from zulip conversation)
* Add documentation
* Add yap...

We don't have a faithful translation of both ParseLikeEcho() and
ParseCmdval().

Refactor to use...

- Run with ysh in spec/ysh-builtins
- shopt -s ysh:all is like running with bin/ysh, which is ...

Co-authored-by: Andy C <[email protected]>

- Notes on TYG8 -- I think there is probably someting to share.
- ASDL is an IR for Oils. And...

We were only printing each value once, where as printing JSON trees can
print values multiple ti...

We're doing reference counting, not cycle detection! In both JSON and
ASDL.

[ASDL] Fix type an...

Still need to do generated C++. I guess that requires a Dict, or maybe
we use something like my...

Also make notes about ref counting, which we can use for:

- = operator
- pp line (x) - since it...

These are valid in J8 Notation (TYG8), but not JSON or JSON8.

Unifying ASDL pretty printing and J8 Notation with "TYG8" (typed graph).

I recently did a bit o...

Technically we could have two different CanOmitQuotes(), but this
doesn't really hurt.

The set ...

ASDL now depends on J8 notation for pretty printing, which depends on
the Python 2 extension fas...

[doc] Release index now links to Oils reference

Wow that is allowed in C, and very unsafe

[spec/xtrace] Adjust stderr for quoting

[doc/ref] Po...

[cleanup] remove parse_tea shopt

Based on a survey, the only pages that don't have this are:

- release-{index,quality}.html - I ...

We can't just refer to the member because it needs to be computed for
small string optimization.

* Create spec tests
* Handle tricky cases where new_len or old_len are zero
* Update doc/ref
...

This is bug #1775.

It was unclear whether we were supporting arbitrary indices from a YSH
user,...

We can have these 3 behaviors depending on whether ^ is present and
whether re.MULTILINE ("m" in...