Add new uber-simple vendor/jsonx.jl for true no-dependency JSON functionality that can be directly vendored in

Author: quinnj

.github/workflows/vendor-tests.yml

@@ -0,0 +1,48 @@
+name: Vendor Tests
+
+on:
+  push:
+    paths:
+      - 'vendor/**'
+  pull_request:
+    paths:
+      - 'vendor/**'
+
+jobs:
+  test:
+    name: Julia ${{ matrix.version }} - ${{ matrix.os }} - ${{ matrix.arch }}
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        version:
+          - '1' # automatically expands to the latest stable 1.x release of Julia
+          - 'min'
+          - 'pre'
+        os:
+          - ubuntu-latest
+          - windows-latest
+        arch:
+          - x64
+        include:
+          - os: macOS-latest
+            arch: aarch64
+            version: 1
+    
+    steps:
+    - name: Checkout code
+      uses: actions/checkout@v4
+      
+    - name: Setup Julia
+      uses: julia-actions/setup-julia@v2
+      with:
+        version: ${{ matrix.version }}
+        arch: ${{ matrix.arch }}
+        
+    - name: Cache Julia packages
+      uses: julia-actions/cache@v2
+      
+    - name: Run vendor tests
+      run: |
+        cd vendor
+        julia test.jl 

Project.toml

@@ -22,7 +22,7 @@ Arrow = "2.8.0"
 ArrowTypes = "2.2"
 Parsers = "1, 2"
 PrecompileTools = "1"
-StructUtils = "2.1"
+StructUtils = "2.3"
 julia = "1.9"
 
 [extras]

README.md

@@ -71,6 +71,10 @@ JSON.isvalidjson(json)
 JSON.json("test.json", j)
 ```
 
+## Vendor Directory
+
+This package includes a `vendor/` directory containing a simplified, no-dependency JSON parser (`JSONX`) that can be vendored (copied) into other projects. See the [vendor README](vendor/README.md) for details.
+
 ## Contributing and Questions
 
 Contributions are very welcome, as are feature requests and suggestions. Please open an

src/JSON.jl

@@ -31,9 +31,9 @@ export JSONText, StructUtils, @noarg, @defaults, @tags, @choosetype, @nonstruct,
         erri += textwidth(c)
         st = iterate(snippet, i)
     end
-
     snippet = replace(snippet, r"[\b\f\n\r\t]" => " ")
-    caret = repeat(' ', erri + 2) * "^"
+    # we call @invoke here to avoid --trim verify errors
+    caret = @invoke(repeat(" "::String, (erri + 2)::Integer)) * "^"
     msg = """
     invalid JSON at byte position $(pos) (line $line_no) parsing type $T: $error
     $snippet$(error == UnexpectedEOF ? " <EOF>" : "...")
@@ -104,12 +104,10 @@ print(io::IO, obj, indent=nothing) = json(io, obj; pretty=something(indent, 0))
 print(a, indent=nothing) = print(stdout, a, indent)
 @doc (@doc json) print
 
-json(a, indent::Integer) = json(a; pretty=indent)
-
 @compile_workload begin
     x = JSON.parse("{\"a\": 1, \"b\": null, \"c\": true, \"d\": false, \"e\": \"\", \"f\": [1,null,true], \"g\": {\"key\": \"value\"}}")
-    # json = JSON.json(x)
-    # isvalid(json)
+    json = JSON.json(x)
+    isvalid(json)
 end
 
 

src/lazy.jl

@@ -269,8 +269,8 @@ function applyobject(keyvalfunc, x::LazyValues)
     @nextbyte
     b == UInt8('}') && return pos + 1
     while true
-        # applystring returns key as a PtrString
-        key, pos = @inline applystring(nothing, LazyValue(buf, pos, JSONTypes.STRING, getopts(x), false))
+        # parsestring returns key as a PtrString
+        key, pos = @inline parsestring(LazyValue(buf, pos, JSONTypes.STRING, opts, false))
         @nextbyte
         if b != UInt8(':')
             error = ExpectedColon
@@ -455,7 +455,7 @@ StructUtils.keyeq(x::PtrString, y::Symbol) = convert(Symbol, x) == y
 # or not. It allows materialize, _binary, etc. to deal
 # with the string data appropriately without forcing a String allocation
 # PtrString should NEVER be visible to users though!
-function applystring(f, x::LazyValue)
+function parsestring(x::LazyValue)
     buf, pos = getbuf(x), getpos(x)
     len, b = getlength(buf), getbyte(buf, pos)
     if b != UInt8('"')
@@ -483,12 +483,7 @@ function applystring(f, x::LazyValue)
         @nextbyte(false)
     end
     str = PtrString(pointer(buf, spos), pos - spos, escaped)
-    if f === nothing
-        return str, pos + 1
-    else
-        f(str)
-        return pos + 1
-    end
+    return str, pos + 1
 
 @label invalid
     invalid(error, buf, pos, "string")
@@ -519,13 +514,35 @@ macro check_special(special, value)
             i += 1
         end
         if i > length(bytes)
-            valfunc($value)
-            return pos
+            return NumberResult($value), pos
         end
     end)
 end
 
-function applynumber(valfunc, x::LazyValue)
+const INT = 0x00
+const FLOAT = 0x01
+const BIGINT = 0x02
+const BIGFLOAT = 0x03
+const BIG_ZERO = BigInt(0)
+
+struct NumberResult
+    tag::UInt8
+    int::Int64
+    float::Float64
+    bigint::BigInt
+    bigfloat::BigFloat
+    NumberResult(int::Int64) = new(INT, int)
+    NumberResult(float::Float64) = new(FLOAT, Int64(0), float)
+    NumberResult(bigint::BigInt) = new(BIGINT, Int64(0), 0.0, bigint)
+    NumberResult(bigfloat::BigFloat) = new(BIGFLOAT, Int64(0), 0.0, BIG_ZERO, bigfloat)
+end
+
+isint(x::NumberResult) = x.tag == INT
+isfloat(x::NumberResult) = x.tag == FLOAT
+isbigint(x::NumberResult) = x.tag == BIGINT
+isbigfloat(x::NumberResult) = x.tag == BIGFLOAT
+
+@inline function parsenumber(x::LazyValue)
     buf = getbuf(x)
     pos = getpos(x)
     len = getlength(buf)
@@ -610,23 +627,20 @@ function applynumber(valfunc, x::LazyValue)
             # if we overflowed, then let's try BigFloat
             bres = Parsers.xparse2(BigFloat, buf, startpos, len)
             if !Parsers.invalid(bres.code)
-                valfunc(bres.val)
-                return startpos + bres.tlen
+                return NumberResult(bres.val), startpos + bres.tlen
             end
         end
         if Parsers.invalid(res.code)
             error = InvalidNumber
             @goto invalid
         end
-        valfunc(res.val)
-        return startpos + res.tlen
+        return NumberResult(res.val), startpos + res.tlen
     else
         if overflow
-            valfunc(isneg ? -bval : bval)
+            return NumberResult(isneg ? -bval : bval), pos
         else
-            valfunc(isneg ? -val : val)
+            return NumberResult(isneg ? -val : val), pos
         end
-        return pos
     end
 
 @label invalid
@@ -643,9 +657,11 @@ function skip(x::LazyValues)
     elseif T == JSONTypes.ARRAY
         return applyarray((i, v) -> 0, x)
     elseif T == JSONTypes.STRING
-        return applystring(s -> 0, x)
+        _, pos = parsestring(x)
+        return pos
     elseif T == JSONTypes.NUMBER
-        return applynumber(n -> 0, x)
+        _, pos = parsenumber(x)
+        return pos
     elseif T == JSONTypes.TRUE
         return getpos(x) + 4
     elseif T == JSONTypes.FALSE
@@ -716,7 +732,7 @@ function Base.show(io::IO, x::LazyValue)
             show(io, MIME"text/plain"(), la)
         end
     elseif T == JSONTypes.STRING
-        str, _ = applystring(nothing, x)
+        str, _ = parsestring(x)
         Base.print(io, "JSON.LazyValue(", repr(convert(String, str)), ")")
     elseif T == JSONTypes.NULL
         Base.print(io, "JSON.LazyValue(nothing)")