add notes on precedence of unary operators against variable binding

README.md

@@ -78,7 +78,7 @@ docker run -i --rm ghcr.io/itchyny/gojq
 - gojq does not keep the order of object keys. I understand this might cause problems for some scripts but basically, we should not rely on the order of object keys. Due to this limitation, gojq does not have `keys_unsorted` function and `--sort-keys` (`-S`) option. I would implement when ordered map is implemented in the standard library of Go but I'm less motivated.
 - gojq supports arbitrary-precision integer calculation while jq does not; jq loses the precision of large integers when calculation is involved. Note that even with gojq, all mathematical functions, including `floor` and `round`, convert integers to floating-point numbers; only addition, subtraction, multiplication, modulo, and division operators (when divisible) keep the integer precision. To calculate floor division of integers without losing the precision, use `def idivide($n): (. - . % $n) / $n;`. To round down floating-point numbers to integers, use `def ifloor: floor | tostring | tonumber;`, but note that this function does not work with large floating-point numbers and also loses the precision of large integers.
 - gojq fixes various bugs of jq. gojq correctly deletes elements of arrays by `|= empty` ([jq#2051](https://github.com/stedolan/jq/issues/2051)). gojq fixes `try`/`catch` handling ([jq#1859](https://github.com/stedolan/jq/issues/1859), [jq#1885](https://github.com/stedolan/jq/issues/1885), [jq#2140](https://github.com/stedolan/jq/issues/2140)). gojq fixes `nth/2` to output nothing when the count is equal to or larger than the stream size ([jq#1867](https://github.com/stedolan/jq/issues/1867)). gojq consistently counts by characters (not by bytes) in `index`, `rindex`, and `indices` functions; `"１２３４５" | .[index("３"):]` results in `"３４５"` ([jq#1430](https://github.com/stedolan/jq/issues/1430), [jq#1624](https://github.com/stedolan/jq/issues/1624)). gojq handles overlapping occurrence differently in `rindex` and `indices`; `"ababa" | [rindex("aba"), indices("aba")]` results in `[2,[0,2]]` ([jq#2433](https://github.com/stedolan/jq/issues/2433)). gojq supports string indexing; `"abcde"[2]` ([jq#1520](https://github.com/stedolan/jq/issues/1520)). gojq accepts indexing query `.e0` ([jq#1526](https://github.com/stedolan/jq/issues/1526), [jq#1651](https://github.com/stedolan/jq/issues/1651)), and allows `gsub` to handle patterns including `"^"` ([jq#2148](https://github.com/stedolan/jq/issues/2148)). gojq improves variable lexer to allow using keywords for variable names, especially in binding patterns, also disallows spaces after `$` ([jq#526](https://github.com/stedolan/jq/issues/526)). gojq fixes handling files with no newline characters at the end ([jq#2374](https://github.com/stedolan/jq/issues/2374)).
-- gojq truncates down floating-point numbers on indexing (`[0] | .[0.5]` results in `0` not `null`), and slicing (`[0,1,2] | .[0.5:1.5]` results in `[0]` not `[0,1]`). gojq implements `@uri` to escape all the reserved characters defined in RFC 3986, Sec. 2.2 ([jq#1506](https://github.com/stedolan/jq/issues/1506)), and fixes `@base64d` to allow binary string as the decoded string ([jq#1931](https://github.com/stedolan/jq/issues/1931)). gojq improves time formatting and parsing; deals with `%f` in `strftime` and `strptime` ([jq#1409](https://github.com/stedolan/jq/issues/1409)), parses timezone offsets with `fromdate` and `fromdateiso8601` ([jq#1053](https://github.com/stedolan/jq/issues/1053)), supports timezone name/offset with `%Z`/`%z` in `strptime` ([jq#929](https://github.com/stedolan/jq/issues/929), [jq#2195](https://github.com/stedolan/jq/issues/2195)), and looks up correct timezone during daylight saving time on formatting with `%Z` ([jq#1912](https://github.com/stedolan/jq/issues/1912)). gojq supports nanoseconds in date and time functions.
+- gojq truncates down floating-point numbers on indexing (`[0] | .[0.5]` results in `0` not `null`), and slicing (`[0,1,2] | .[0.5:1.5]` results in `[0]` not `[0,1]`). gojq parses unary operators with higher precedence than variable binding (`[-1 as $x | 1,$x]` results in `[1,-1]` not `[-1,-1]`). gojq implements `@uri` to escape all the reserved characters defined in RFC 3986, Sec. 2.2 ([jq#1506](https://github.com/stedolan/jq/issues/1506)), and fixes `@base64d` to allow binary string as the decoded string ([jq#1931](https://github.com/stedolan/jq/issues/1931)). gojq improves time formatting and parsing; deals with `%f` in `strftime` and `strptime` ([jq#1409](https://github.com/stedolan/jq/issues/1409)), parses timezone offsets with `fromdate` and `fromdateiso8601` ([jq#1053](https://github.com/stedolan/jq/issues/1053)), supports timezone name/offset with `%Z`/`%z` in `strptime` ([jq#929](https://github.com/stedolan/jq/issues/929), [jq#2195](https://github.com/stedolan/jq/issues/2195)), and looks up correct timezone during daylight saving time on formatting with `%Z` ([jq#1912](https://github.com/stedolan/jq/issues/1912)). gojq supports nanoseconds in date and time functions.
 - gojq does not support some functions intentionally; `get_jq_origin`, `get_prog_origin`, `get_search_list` (unstable, not listed in jq document), `input_line_number`, `$__loc__` (performance issue), `recurse_down` (deprecated in jq). gojq does not support some flags; `--ascii-output, -a` (performance issue), `--seq` (not used commonly), `--sort-keys, -S` (sorts by default because `map[string]interface{}` does not keep the order), `--unbuffered` (unbuffered by default). gojq does not parse JSON extensions supported by jq; `NaN`, `Infinity`, and `[000]`. gojq normalizes floating-point numbers to fit to double-precision floating-point numbers. gojq does not support some regular expression flags (regular expression engine differences). gojq does not support BOM (`encoding/json` does not support this). gojq disallows using keywords for function names (`def true: .; true` is a confusing query), and module name prefixes in function declarations (using module prefixes like `def m::f: .;` is undocumented).
 - gojq supports reading from YAML input (`--yaml-input`) while jq does not. gojq also supports YAML output (`--yaml-output`).
 

cli/test.yaml

@@ -837,13 +837,21 @@
     [-1,-2,-3,-4722366482869645213696]
     [1,2,3,4722366482869645213696]
 
-- name: unary operators against string
+- name: unary operator against string
   args:
     - '(-.)'
   input: '"abcdeabcdeabcdeabcdeabcdeabcdeabcdeabcdeabcdeabcde"'
   error: |
     cannot negate: string ("abcdeabcdeabcdeabcdeabcd ...")
 
+- name: unary operator with variable binding
+  args:
+    - '-1 as $x | 1, $x'
+  input: 'null'
+  expected: |
+    1
+    -1
+
 - name: object construction
   args:
     - -c