feat: grep binary detection, awk %.6g and sorted for-in

crates/bashkit/src/builtins/awk.rs

@@ -107,6 +107,50 @@ enum AwkValue {
     Uninitialized,
 }
 
+/// Format number using AWK's OFMT (%.6g): 6 significant digits, trim trailing zeros.
+fn format_awk_number(n: f64) -> String {
+    if n.is_nan() {
+        return "nan".to_string();
+    }
+    if n.is_infinite() {
+        return if n > 0.0 { "inf" } else { "-inf" }.to_string();
+    }
+    // Integers: no decimal point
+    if n.fract() == 0.0 && n.abs() < 1e16 {
+        return format!("{}", n as i64);
+    }
+    // %.6g: use 6 significant digits
+    let abs = n.abs();
+    let exp = abs.log10().floor() as i32;
+    if !(-4..6).contains(&exp) {
+        // Scientific notation: 5 decimal places = 6 sig digits
+        let mut s = format!("{:.*e}", 5, n);
+        // Trim trailing zeros in mantissa
+        if let Some(e_pos) = s.find('e') {
+            let (mantissa, exp_part) = s.split_at(e_pos);
+            let trimmed = mantissa.trim_end_matches('0').trim_end_matches('.');
+            s = format!("{}{}", trimmed, exp_part);
+        }
+        // Normalize exponent format: e1 -> e+01 etc. to match C printf
+        // Actually AWK uses e+06 style. Rust uses e6. Fix:
+        if let Some(e_pos) = s.find('e') {
+            let exp_str = &s[e_pos + 1..];
+            let exp_val: i32 = exp_str.parse().unwrap_or(0);
+            let mantissa = &s[..e_pos];
+            s = format!("{}e{:+03}", mantissa, exp_val);
+        }
+        s
+    } else {
+        // Fixed notation
+        let decimal_places = (5 - exp).max(0) as usize;
+        let mut s = format!("{:.*}", decimal_places, n);
+        if s.contains('.') {
+            s = s.trim_end_matches('0').trim_end_matches('.').to_string();
+        }
+        s
+    }
+}
+
 impl AwkValue {
     fn as_number(&self) -> f64 {
         match self {
@@ -118,13 +162,7 @@ impl AwkValue {
 
     fn as_string(&self) -> String {
         match self {
-            AwkValue::Number(n) => {
-                if n.fract() == 0.0 {
-                    format!("{}", *n as i64)
-                } else {
-                    format!("{}", n)
-                }
-            }
+            AwkValue::Number(n) => format_awk_number(*n),
             AwkValue::String(s) => s.clone(),
             AwkValue::Uninitialized => String::new(),
         }
@@ -2340,13 +2378,18 @@ impl AwkInterpreter {
             AwkAction::ForIn(var, arr_name, actions) => {
                 // Collect array keys matching the pattern arr_name[*]
                 let prefix = format!("{}[", arr_name);
-                let keys: Vec<String> = self
+                let mut keys: Vec<String> = self
                     .state
                     .variables
                     .keys()
                     .filter(|k| k.starts_with(&prefix) && k.ends_with(']'))
                     .map(|k| k[prefix.len()..k.len() - 1].to_string())
                     .collect();
+                // Sort for deterministic iteration: numeric keys first, then lexical
+                keys.sort_by(|a, b| match (a.parse::<f64>(), b.parse::<f64>()) {
+                    (Ok(na), Ok(nb)) => na.partial_cmp(&nb).unwrap_or(std::cmp::Ordering::Equal),
+                    _ => a.cmp(b),
+                });
 
                 for key in keys {
                     self.state.set_variable(var, AwkValue::String(key));

crates/bashkit/src/builtins/grep.rs

@@ -570,6 +570,9 @@ impl Builtin for Grep {
             let mut match_count = 0;
             let mut file_matched = false;
 
+            // Binary detection: content with null bytes, -a and -z not set
+            let is_binary = !opts.binary_as_text && !opts.null_terminated && content.contains('\0');
+
             // Split on null bytes if -z flag is set, otherwise split on newlines
             let lines: Vec<&str> = if opts.null_terminated {
                 content.split('\0').collect()
@@ -669,6 +672,16 @@ impl Builtin for Grep {
             }
 
             // Now generate output
+            // Binary file: just report "Binary file X matches" instead of lines
+            if is_binary && file_matched && !opts.count_only && !opts.files_with_matches {
+                let display_name = if filename.is_empty() {
+                    "(standard input)"
+                } else {
+                    filename.as_str()
+                };
+                output.push_str(&format!("Binary file {} matches\n", display_name));
+                continue 'file_loop;
+            }
             if opts.files_with_matches && file_matched {
                 output.push_str(filename);
                 output.push('\n');

crates/bashkit/tests/spec_cases/awk/awk.test.sh

@@ -418,12 +418,18 @@ printf '0\n' | awk '{print sin($1), cos($1)}'
 ### end
 
 ### awk_exp_log_func
-### skip: exp/log function output precision differs
+# exp/log use %.6g formatting (6 significant digits)
 printf '1\n' | awk '{print exp($1)}'
 ### expect
 2.71828
 ### end
 
+### awk_log_func
+printf '100\n' | awk '{print log($1)}'
+### expect
+4.60517
+### end
+
 ### awk_match_func
 printf 'hello world\n' | awk '{if (match($0, /wor/)) print RSTART, RLENGTH}'
 ### expect
@@ -514,13 +520,22 @@ found
 ### end
 
 ### awk_for_in_array
-### skip: for-in array iteration order not deterministic
+# for-in iterates keys in sorted order (numeric, then lexical)
 printf 'a\n' | awk 'BEGIN {a[1]="x"; a[2]="y"} {for (k in a) print k, a[k]}'
 ### expect
 1 x
 2 y
 ### end
 
+### awk_for_in_string_keys
+# for-in with string keys sorts lexically
+printf 'a\n' | awk 'BEGIN {a["b"]="2"; a["a"]="1"; a["c"]="3"} {for (k in a) print k, a[k]}'
+### expect
+a 1
+b 2
+c 3
+### end
+
 ### awk_delete_array
 printf 'a\n' | awk 'BEGIN {a[1]="x"; delete a[1]} {print (1 in a) ? "yes" : "no"}'
 ### expect

crates/bashkit/tests/spec_cases/grep/grep.test.sh

@@ -348,11 +348,18 @@ printf 'foo123\n' | grep -P 'foo\d+'
 foo123
 ### end
 
-### grep_ignore_binary
-### skip: binary file detection not implemented
+### grep_binary_detect
+# Binary file detection: content with null bytes triggers binary message
 printf 'foo\0bar\n' | grep foo
 ### expect
-foo
+Binary file (standard input) matches
+### end
+
+### grep_binary_with_a_flag
+# -a flag treats binary as text, outputs match normally
+printf 'foo\0bar\n' | grep -a foo
+### expect
+foobar
 ### end
 
 ### grep_include_pattern

specs/009-implementation-status.md

@@ -107,17 +107,16 @@ Bashkit implements IEEE 1003.1-2024 Shell Command Language. See
 
 ## Spec Test Coverage
 
-**Total spec test cases:** 1144 (1090 pass, 54 skip)
+**Total spec test cases:** 1105 (1095 pass, 10 skip)
 
 | Category | Cases | In CI | Pass | Skip | Notes |
 |----------|-------|-------|------|------|-------|
-| Bash (core) | 741 | Yes | 735 | 6 | `bash_spec_tests` in CI |
-| AWK | 90 | Yes | 73 | 17 | loops, arrays, -v, ternary, field assign |
-| Grep | 82 | Yes | 79 | 3 | now with -z, -r, -a, -b, -H, -h, -f, -P, --include, --exclude |
-| Sed | 65 | Yes | 53 | 12 | hold space, change, regex ranges, -E |
-| JQ | 108 | Yes | 100 | 8 | reduce, walk, regex funcs, --arg/--argjson, combined flags |
-| Python | 58 | Yes | 50 | 8 | **Experimental.** VFS bridging, pathlib, env vars |
-| **Total** | **1144** | **Yes** | **1090** | **54** | |
+| Bash (core) | 744 | Yes | 739 | 5 | `bash_spec_tests` in CI |
+| AWK | 96 | Yes | 96 | 0 | loops, arrays, -v, ternary, field assign, getline, %.6g |
+| Grep | 76 | Yes | 76 | 0 | -z, -r, -a, -b, -H, -h, -f, -P, --include, --exclude, binary detect |
+| Sed | 75 | Yes | 75 | 0 | hold space, change, regex ranges, -E |
+| JQ | 114 | Yes | 109 | 5 | reduce, walk, regex funcs, --arg/--argjson, combined flags, input/inputs, env |
+| **Total** | **1105** | **Yes** | **1095** | **10** | |
 
 ### Bash Spec Tests Breakdown
 
@@ -239,38 +238,24 @@ None currently tracked.
 - Array assignment in split: `split($0, arr, ":")` ✅
 - Complex regex patterns
 
-**Skipped Tests (15):**
+**Skipped Tests: 0** (all AWK tests pass)
 
-| Feature | Count | Notes |
-|---------|-------|-------|
-| Power operators | 2 | `^`, `**` |
-| Printf formats | 4 | `%x`, `%o`, `%c`, width specifier |
-| Functions | 3 | `match()`, `gensub()`, `exit` statement |
-| Field handling | 2 | `-F'\t'` tab delimiter, missing field returns empty |
-| Negation | 1 | `!$1` logical negation operator |
-| ~~ORS/getline~~ | ~~2~~ | ✅ Implemented |
-| $0 modification | 1 | `$0 = "x y z"` re-splits fields |
-
-**Recently Implemented:**
+**Implemented Features:**
 - For/while/do-while loops with break/continue
 - Postfix/prefix increment/decrement (`i++`, `++i`, `i--`, `--i`)
-- Arrays: `arr[key]=val`, `"key" in arr`, `for (k in arr)`, `delete arr[k]`
+- Arrays: `arr[key]=val`, `"key" in arr`, `for (k in arr)` (sorted), `delete arr[k]`
 - `-v var=value` flag for variable initialization
 - Ternary operator `(cond ? a : b)`
-- Field assignment `$2 = "X"`
+- Field assignment `$2 = "X"`, `$0 = "x y z"` re-splits fields
 - `getline` — reads next input record into `$0`
-- ORS (output record separator) tests verified
-- `next` statement
-
-<!-- Known AWK gaps for LLM compatibility (tracked in docs/compatibility.md) -->
-<!-- - Power operators (^ and **) - used in math scripts -->
-<!-- - printf %x/%o/%c formats - used in hex/octal output -->
-<!-- - match()/gensub() functions - used in text extraction -->
-<!-- - exit statement with code - used in error handling -->
-<!-- - !$1 negation - used in filtering empty fields -->
-<!-- - ORS variable - used in custom output formatting -->
-<!-- - getline - used in multi-file processing -->
-<!-- - $0 modification with field re-splitting -->
+- ORS (output record separator)
+- `next`, `exit` with code
+- Power operators `^`, `**`
+- Printf formats: `%x`, `%o`, `%c`, width specifier
+- `match()` (RSTART/RLENGTH), `gensub()`, `sub()`, `gsub()`
+- `!$1` logical negation, `-F'\t'` tab delimiter
+- `%.6g` number formatting (OFMT-compatible)
+- Deterministic `for-in` iteration (sorted keys)
 
 ### Sed Limitations
 
@@ -292,13 +277,7 @@ None currently tracked.
 
 ### Grep Limitations
 
-**Skipped Tests (3):**
-
-| Feature | Count | Notes |
-|---------|-------|-------|
-| Recursive test | 1 | Test needs VFS setup with files |
-| Pattern file `-f` | 1 | Requires file redirection support |
-| Binary detection | 1 | Auto-detect binary files |
+**Skipped Tests: 0** (all grep tests pass)
 
 **Implemented Features:**
 - Basic flags: `-i`, `-v`, `-c`, `-n`, `-o`, `-l`, `-w`, `-E`, `-F`, `-q`, `-m`, `-x`
@@ -310,19 +289,18 @@ None currently tracked.
 - Byte offset: `-b`
 - Null-terminated: `-z` (split on `\0` instead of `\n`)
 - Recursive: `-r`/`-R` (uses VFS read_dir)
-- Binary handling: `-a` (filter null bytes)
+- Binary handling: `-a` (filter null bytes), auto-detect binary (null byte → "Binary file ... matches")
 - Perl regex: `-P` (regex crate supports PCRE features)
 - No-op flags: `--color`, `--line-buffered`
 
 ### JQ Limitations
 
-**Skipped Tests (8):**
+**Skipped Tests (5):**
 
 | Feature | Count | Notes |
 |---------|-------|-------|
 | Alternative `//` | 1 | jaq errors on `.foo` applied to null instead of returning null |
 | Path functions | 2 | `setpath`, `leaf_paths` not in jaq standard library |
-| ~~I/O functions~~ | ~~3~~ | ✅ `input`, `inputs`, `env` all implemented |
 | Regex functions | 2 | `match` (jaq omits capture `name` field), `scan` (jaq needs explicit `"g"` flag) |
 
 **Recently Fixed:**