Change in behavoir of query argument of fparse between version 0.1.1 and 0.1.2

Question

Change in behavoir of query argument of fparse between version 0.1.1 and 0.1.2

mem48 opened this issue 4 years ago · comments

I'm updated my package, and it seems the new version of RcppSimdJson has changed whether the query argument needs a leading slash.

For example:

json_string = '{"a":[[1,2,3],["a","b","c"]],"b":[[4,5,6],["d","e","f"]]}'

# Works in 0.1.1, but doesn't work in 0.1.2
RcppSimdJson::fparse(json_string, query = "a") 

# Doesn't work in 0.1.1, but works in 0.1.2
RcppSimdJson::fparse(json_string, query = "/a")

Is this an intentional permanent change? In which case I need to update my package to require version >= 0.1.2.

Dirk Eddelbuettel · Answer 1 · Fri Oct 23 2020 22:30:27 GMT+0800 (China Standard Time)

Correct. That comes from upstream in 0.5.0 and there may be another change in 0.6.0 (which I am currently updating to and where we saw a few more (miniscule) changes in error behavior).

You will have to update your code for the leading slash.

/cc @lemire

Malcolm Morgan · Answer 2 · Fri Oct 23 2020 23:47:24 GMT+0800 (China Standard Time)

Thanks for the clarification.

I've been getting an error on CRAN after updating my package, so I wanted to check that I really had to enforce using version 0.1.2

Version: 0.3.0
Check: package dependencies
Result: ERROR
    Package required but not available: ‘RcppSimdJson’
    
    See section ‘The DESCRIPTION file’ in the ‘Writing R Extensions’
    manual.
Flavor: r-patched-solaris-x86

Dirk Eddelbuettel · Answer 3 · Sat Oct 24 2020 00:41:55 GMT+0800 (China Standard Time)

That may be between you and CRAN. There can be occassional race conditions on the builders, this comes up from time to time on the (excellent) r-package-devel list and the recommendation (by Uwe Ligges) is usually to just resubmit (if your're sure it is not you).

Looking at https://cloud.r-project.org/web/packages/RcppSimdJson/index.html does not suggest any binaries are missing.

So on balance I am going to close this. Please feel free to reopen if I overlooked anything.

Daniel Lemire · Answer 4 · Sat Oct 24 2020 01:16:32 GMT+0800 (China Standard Time)

At some point, some of us misread the specifications (RFC) and this lead to an improper syntax parsing. I am very sorry about this, but... yes, the leading slashes are needed as per JSON Pointer.

Our own syntax was self-consistent, but it did not match anything that was formally published. We were silently innovating (without realizing it).

Malcolm Morgan · Answer 5 · Sat Oct 24 2020 04:25:42 GMT+0800 (China Standard Time)

Thanks for the speedy and helpful response, and for writing an amazing package.

Malcolm Morgan · Answer 6 · Mon Oct 26 2020 19:56:48 GMT+0800 (China Standard Time)

Hi @eddelbuettel

Looking at https://cloud.r-project.org/web/packages/RcppSimdJson/index.html I see that the windows binaries for r-oldrel are not available.

I can confirm that I can't install rcppsimdjson on Windows with R 3.6 and it won't compile from source either.

Dirk Eddelbuettel · Answer 7 · Mon Oct 26 2020 19:58:38 GMT+0800 (China Standard Time)

Yes that is the way it goes with compilers that are too old.

R 4.0.0 was released six months ago now following the usual extensive testing that effectively takes a year during which r-devel ripens, and has by now seen three more patch releases. If you want to benefit from "modern C++" alongside it you need to upgrade and will also get Rtools4 in the process.