option to allow trailing characters while parsing

[tanmaykm]

This exposes the existing internal isroot parameter of _lazy as a keyword argument on JSON.lazy (and by extension JSON.parse).

By default, isroot=true, which means the parser expects the entire buffer to be a single valid JSON value — trailing characters after the root value will raise an error. Setting isroot=false parses only the first JSON value from the buffer and silently ignores any trailing characters.

Useful for:

• Parsing buffers that contain multiple concatenated JSON objects without a delimiter, e.g. {"a":1}{"b":2}
• Parsing JSON followed by non-JSON content, e.g. {"a":1} : some annotation...
• Restoring the pre-1.x behavior of this package

julia> JSON.parse("{\"hello\": \"world\"} extra stuff", isroot=false)
  JSON.Object with 1 entry:
  "hello" => "world"

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.26%. Comparing base (f4fbb5a) to head (5b0ddcd).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #439   +/-   ##
=======================================
  Coverage   90.26%   90.26%           
=======================================
  Files           7        7           
  Lines        1366     1366           
=======================================
  Hits         1233     1233           
  Misses        133      133           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[quinnj]

I don't love introducing a new keyword argument/option for this, especially when we have the isroot property right there. I think I'd prefer allowing passing isroot=false as a keyword arg that would get passed down and then we'd have:

function lazy(buf::Union{AbstractVector{UInt8}, AbstractString}; isroot::Bool=true, kw...)
    if !applicable(pointer, buf, 1) || (buf isa AbstractVector{UInt8} && !isone(only(strides(buf))))
        if buf isa AbstractString
            buf = String(buf)
        else
            buf = Vector{UInt8}(buf)
        end
    end
    len = getlength(buf)
    if len == 0
        error = UnexpectedEOF
        pos = 0
        @goto invalid
    end
    pos = 1
    # detect and error on UTF-16LE BOM
    if len >= 2 && getbyte(buf, pos) == 0xff && getbyte(buf, pos + 1) == 0xfe
        error = InvalidUTF16
        @goto invalid
    end
    # detect and error on UTF-16BE BOM
    if len >= 2 && getbyte(buf, pos) == 0xfe && getbyte(buf, pos + 1) == 0xff
        error = InvalidUTF16
        @goto invalid
    end
    # detect and ignore UTF-8 BOM
    pos = (len >= 3 && getbyte(buf, pos) == 0xef && getbyte(buf, pos + 1) == 0xbb && getbyte(buf, pos + 2) == 0xbf) ? pos + 3 : pos
    @nextbyte
    return _lazy(buf, pos, len, b, LazyOptions(; kw...), isroot)

@label invalid
    invalid(error, buf, pos, Any)
end

the main differences being that we "capture" the isroot::Bool=true keyword arg (so it isn't passed down to _lazy and then we construct the LazyValue w/ user-provided isroot.

[tanmaykm]

Thanks @quinnj , this is much cleaner. I have updated the PR with your suggestion.

[quinnj]

Looking pretty good; will you also add this into the docs for JSON.parse in the parse.jl file? (we have several of the JSON.lazy keyword args repeated there). Then if you include a minor version bump, we can merge and release.

[tanmaykm]

Done. Thanks!