Output formats and piping

Every command takes --format. Pick json when something will parse it, table when you will read it, and ids when the output feeds the next command.

lazydap stack --format json | jq '.frames[0].line'

With no --format, lazydap chooses: table when stdout is a terminal, json everywhere else. So a pipeline gets JSON without asking, and a human gets columns without asking.

The five

json — the contract

One object or array. The schema is stable and breaking it costs a decision-log entry.

$ lazydap variables --reference 1003 --format json
{
  "variables": [
    { "name": "n",   "type_name": "int", "value": "10", "variables_reference": 0 },
    { "name": "sum", "type_name": "int", "value": "21", "variables_reference": 0 },
    { "name": "i",   "type_name": "int", "value": "7",  "variables_reference": 0 }
  ]
}

table — not a contract

$ lazydap variables --reference 1003 --format table
NAME  VALUE  TYPE  REFERENCE
n     10     int   0
sum   21     int   0
i     7      int   0

Column order and widths change whenever they read badly. Do not parse this. If you are reaching for awk here, you want --format json or --format csv.

jsonl — one object per line

For streams and while read. Arrays come out as one element per line rather than one blob:

$ lazydap stack --format jsonl
{"column":16,"id":1010,"line":6,"name":"total","source":{"name":"hello.c","path":"/Users/you/lazydap-demo/hello.c"}}
{"column":18,"id":1011,"line":14,"name":"main","source":{"name":"hello.c","path":"/Users/you/lazydap-demo/hello.c"}}
{"column":0,"id":1012,"line":1751,"name":"start","source":{"name":"@start","source_reference":1000}}

csv — for spreadsheets and cut

$ lazydap stack --format csv
frame,name,source,line
1007,total,/Users/you/lazydap-demo/hello.c,6
1008,main,/Users/you/lazydap-demo/hello.c,14
1009,start,@start,1751

A flattened view, so nested fields are dropped rather than escaped. Use json when you need all of it.

ids — bare ids for xargs

$ lazydap break --list --format ids
1

One id per line, nothing else. This is the format that makes commands compose.

Piping between commands

--id is repeatable and reads exactly what --format ids writes:

lazydap break --list --format ids | xargs -I{} lazydap break --remove --id {}

Preview it first. Every mutation takes --dry-run and reports the same selection the real run would act on:

lazydap break --remove --all --dry-run --format json
lazydap break --remove --all

xargs -r is GNU-only and does nothing on macOS. When an empty list is possible, guard it:

ids=$(lazydap break --list --format ids)
[ -n "$ids" ] && printf '%s\n' "$ids" | xargs -I{} lazydap break --remove --id {}

Reading a --wait reply with jq

result=$(lazydap continue --wait --format json)

printf '%s' "$result" | jq -r .state
printf '%s' "$result" | jq -r '.frame | "\(.source.name):\(.line) in \(.name)"'
printf '%s' "$result" | jq -r '.captured_output[] | select(.category=="stdout") | .output'

The last one is the program’s own output, separated from the adapter’s console chatter.

Errors

Errors go to stderr as JSON, and stdout stays empty. So a pipeline that parses stdout does not get an error object where it expected data:

$ lazydap variables --ref 1006 --format json
{"details":{"kind":"UnknownArgument"},"error":"UsageError","message":"error: unexpected argument '--ref' found\n\n  tip: a similar argument exists: '--reference'\n\nUsage: lazydap variables --reference <REFERENCE>\n\nFor more information, try '--help'."}
$ echo $?
2

The exit code is the signal to branch on. Errors and exit codes lists all of them.

See also

• JSON output — field-by-field schemas
• lazydap break — selection flags in full
• Debug with an agent — the same formats, from a model