Discord Username / User ID
uisupersaiyan3
What does this improvement do?
The only machine-readable output today is --output json, which for the paginated list surfaces emits a single buffered, pretty-printed array (out.print({ runs, nextCursor }, ...) in runResultHistory at src/commands/test.ts:3735, test runList at :419, project runList at src/commands/project.ts:47). That shape is awkward for the two most common scripting consumers: spreadsheets and awk (must hand-parse JSON), and line-oriented pipelines (jq or while read must buffer and slice the whole array). This adds two modes to the existing --output enum. --output csv emits an RFC 4180 table: a header row, then one row per item, with columns derived from the response type so they cannot drift. --output ndjson emits one compact JSON object per line with no wrapping array, streamed as each page arrives, so ... --output ndjson | jq -c 'select(.status=="failed")' works without buffering. Both apply to the list commands only (test result --history, project list, test list); a single-object command given --output csv or --output ndjson exits 5 with a clear message. It drops straight into shell CI and data pipelines.
Details / implementation notes
Extend OutputMode (src/lib/output.ts:1) to 'json' | 'text' | 'csv' | 'ndjson' and widen isOutputMode (:12); because --output is validated at one choke point (src/index.ts:128, :140) the new modes stay uniform across command groups. Reuse (do not re-implement) the existing render call sites: runResultHistory (src/commands/test.ts:3735, table via renderRunHistoryTable :3860), test runList (:419, renderTestListText :473), project runList (src/commands/project.ts:47), by adding csv and ndjson branches alongside the current json/text branch in each. Row columns come from the source-of-truth types (RunHistoryItem, CliTest, CliProject), never hand-listed. csv quoting follows RFC 4180 (fields containing , " or a newline are wrapped, " is doubled; null/absent renders as an empty field). ndjson writes one line per row incrementally across the existing cursor loop and routes pagination metadata (nextCursor, short-page hints) to stderr so stdout stays pure ndjson. --output json and --output text remain byte-identical; exit codes are unchanged except exit 5 when csv or ndjson is used on a non-list command. No new dependency. Tests: csv quoting edge cases, multi-page accumulation, ndjson line shape, stdout/stderr separation, non-list rejection exit 5, json/text unchanged. PR to follow once assigned.
Confirmations
Discord Username / User ID
uisupersaiyan3
What does this improvement do?
The only machine-readable output today is
--output json, which for the paginated list surfaces emits a single buffered, pretty-printed array (out.print({ runs, nextCursor }, ...)inrunResultHistoryatsrc/commands/test.ts:3735, testrunListat:419, projectrunListatsrc/commands/project.ts:47). That shape is awkward for the two most common scripting consumers: spreadsheets andawk(must hand-parse JSON), and line-oriented pipelines (jqorwhile readmust buffer and slice the whole array). This adds two modes to the existing--outputenum.--output csvemits an RFC 4180 table: a header row, then one row per item, with columns derived from the response type so they cannot drift.--output ndjsonemits one compact JSON object per line with no wrapping array, streamed as each page arrives, so... --output ndjson | jq -c 'select(.status=="failed")'works without buffering. Both apply to the list commands only (test result --history,project list,test list); a single-object command given--output csvor--output ndjsonexits 5 with a clear message. It drops straight into shell CI and data pipelines.Details / implementation notes
Extend
OutputMode(src/lib/output.ts:1) to'json' | 'text' | 'csv' | 'ndjson'and widenisOutputMode(:12); because--outputis validated at one choke point (src/index.ts:128,:140) the new modes stay uniform across command groups. Reuse (do not re-implement) the existing render call sites:runResultHistory(src/commands/test.ts:3735, table viarenderRunHistoryTable:3860), testrunList(:419,renderTestListText:473), projectrunList(src/commands/project.ts:47), by adding csv and ndjson branches alongside the current json/text branch in each. Row columns come from the source-of-truth types (RunHistoryItem,CliTest,CliProject), never hand-listed. csv quoting follows RFC 4180 (fields containing,"or a newline are wrapped,"is doubled;null/absent renders as an empty field). ndjson writes one line per row incrementally across the existing cursor loop and routes pagination metadata (nextCursor, short-page hints) to stderr so stdout stays pure ndjson.--output jsonand--output textremain byte-identical; exit codes are unchanged except exit 5 when csv or ndjson is used on a non-list command. No new dependency. Tests: csv quoting edge cases, multi-page accumulation, ndjson line shape, stdout/stderr separation, non-list rejection exit 5, json/text unchanged. PR to follow once assigned.Confirmations