Add JSON Lines (JSONL) input mode

[Copilot]

Adds jsonl as a new input mode, enabling GoAWK to process JSON Lines format where each line is a JSON value. Complements the existing CSV/TSV modes.

New mode: INPUTMODE="jsonl" / -i jsonl

JSON arrays → positional fields $1, $2, ...

# Input: ["Gilbert", "2013", 24, true]
BEGIN { INPUTMODE="jsonl" } { print $1, $2, $3, $4 }
# → Gilbert 2013 24 1

JSON objects → named fields via @"key", also accessible as $1, $2, ... in document key order; FIELDS array updated per record

# Input: {"version":"1.1","host":"example.org","short_message":"A log message"}
BEGIN { INPUTMODE="jsonl" } { print @"host", @"short_message" }
# → example.org A log message

Nested objects and arrays are flattened using dot notation for objects and numeric indexes for arrays:

# Input: {"one": 1, "four": [1,2,3,4], "five": {"alpha": ["fo", "fum"], "beta": {"hey": "How's tricks?"}}}
BEGIN { INPUTMODE="jsonl" } { print @"four.0", @"five.alpha.1", @"five.beta.hey" }
# → 1 fum How's tricks?

Accessing a non-scalar key (e.g. @"four" where four is an array) returns "". Positional fields $1, $2, ... and the FIELDS array reflect all flattened scalar leaves in document order.

Type conversions

JSON	AWK
true	"1"
false	"0"
null	""
number	numeric string
string	string (unescaped)

Implementation

• interp/jsonl.go (new): jsonlSplitter (mirrors csvSplitter — pre-parses each line before action bodies run, so FIELDS and per-record state are ready), parseJSONLineToFields(), flattenObject() / flattenArray() / flattenJSONValue() for recursive flattening, and JSON→AWK conversion helpers
• interp/interp.go: JSONLMode IOMode = 3; getFieldByName() calls ensureFields() first in JSONL mode (per-record object keys, unlike CSV's fixed header); updated parseInputMode() / inputModeString()
• interp/io.go: newScanner() uses jsonlSplitter for JSONLMode; ensureFields() re-parses from $0 on reassignment; setFieldNames(nil) correctly clears FIELDS for array records
• Empty lines are skipped; $0 holds the raw JSON line; $0 reassignment re-parses as JSON

[benhoyt]

@copilot thanks. For nested objects and stats, I'd much rather these were flattened to @"foo.nested" or @"foo.0". Can you please try that?

[Copilot]

@copilot thanks. For nested objects and stats, I'd much rather these were flattened to @"foo.nested" or @"foo.0". Can you please try that?

Done in 9612689. Nested JSON values are now flattened recursively using dot notation:

• Nested object key foo.bar → @"foo.bar"
• Array element at index i under key foo → @"foo.0", @"foo.1", etc.
• This works at any nesting depth: @"five.alpha.1", @"five.beta.hey", etc.

For example:

{"one": 1, "four": [1,2,3,4], "five": {"alpha": ["fo", "fum"], "beta": {"hey": "How's tricks?"}}}

gives @"four.0" → 1, @"five.alpha.1" → fum, @"five.beta.hey" → How's tricks?

Accessing @"foo" where foo is a non-scalar key returns "" (empty string, not an error). Positional fields $1, $2, ... and FIELDS reflect all flattened scalar leaves in document order.