feat(drive): add +search shortcut with flat filter flags (#658)

Expose doc_wiki/search v2 under the drive domain via explicit flags
(--query, --edited-since, --commented-since, --opened-since,
--created-since, --mine, --creator-ids, --doc-types, --folder-tokens,
--space-ids, ...) instead of a nested JSON filter, so natural-language
queries from AI agents map 1:1 to discrete flags.

Time handling:
- my_edit_time and my_comment_time are snapped to the hour (floor/ceil)
  with a stderr notice, since those fields are aggregated at hour
  granularity server-side. create_time passes through as-is.
- open_time has a server-side 3-month cap per request. When
  --opened-since / --opened-until span exceeds 90 days, the CLI narrows
  the request to the most recent 90-day slice and emits a stderr notice
  listing every remaining slice's --opened-* values so the agent can
  re-invoke for older ranges. Spans over 365 days are rejected up front
  to bound runaway slicing.

Flag ergonomics:
- --doc-types accepts mixed case; values are normalized to upper case
  before validation and before being sent to the server.
- --sort default is translated to the server enum DEFAULT_TYPE (every
  other sort value upper-cases 1:1).

Error hints:
- Lark code 99992351 (referenced open_id outside the app's contact
  visibility) is enriched with a +search-specific hint that
  distinguishes API scope from contact visibility and points at
  --creator-ids / --sharer-ids as the likely source.

Skill docs:
- new reference at skills/lark-drive/references/lark-drive-search.md,
  including the open_time slicing protocol and the paginate-within-
  slice-before-switching agent playbook.
- lark-drive/SKILL.md routes resource-discovery to drive +search.
- lark-doc/SKILL.md and lark-doc-search.md mark docs +search as
  deprecated and point users at drive +search.

Change-Id: I36d620045809b448446d4fdbdfa923b05794da19
This commit is contained in:
liujinkun2025
2026-04-25 16:22:35 +08:00
committed by GitHub
parent 2e7a11a8e8
commit aa48d70d7a
9 changed files with 2358 additions and 5 deletions

View File

@@ -0,0 +1,806 @@
// Copyright (c) 2026 Lark Technologies Pte. Ltd.
// SPDX-License-Identifier: MIT
package drive
import (
"context"
"encoding/json"
"errors"
"fmt"
"io"
"math"
"regexp"
"strconv"
"strings"
"time"
"github.com/larksuite/cli/internal/output"
"github.com/larksuite/cli/shortcuts/common"
)
// driveSearchErrUserNotVisible is the Lark service code returned by
// doc_wiki/search when an open_id referenced in --creator-ids / --sharer-ids
// falls outside the app's user-visibility scope (different from the
// search:docs:read API scope).
const driveSearchErrUserNotVisible = 99992351
// open_time has a server-side cap of 3 months per request. Rather than
// reject or silently clamp, we narrow this request to the most recent
// 3-month slice and list the remaining slices in a stderr notice so the
// agent can re-invoke for older ranges.
const (
driveSearchSliceDays = 90 // one slice = server-side 3-month cap
driveSearchMaxOpenedSpanDays = 365 // hard cap: reject --opened-* spans beyond ~1 year
)
var driveSearchSortValues = []string{
"default",
"edit_time",
"edit_time_asc",
"open_time",
"create_time",
}
var driveSearchDocTypeSet = map[string]struct{}{
"DOC": {}, "SHEET": {}, "BITABLE": {}, "MINDNOTE": {}, "FILE": {},
"WIKI": {}, "DOCX": {}, "FOLDER": {}, "CATALOG": {}, "SLIDES": {}, "SHORTCUT": {},
}
// driveSearchHourAggregatedFields lists filter keys the server aggregates at
// hour granularity. We pre-snap start/end and emit a stderr notice so callers
// see what was sent and why.
var driveSearchHourAggregatedFields = map[string]struct{}{
"my_edit_time": {},
"my_comment_time": {},
}
// Server caps list filters at 20 entries each. We reject above-cap input
// locally so users and agents get a named-flag error instead of an opaque
// server-side failure or truncated result.
const (
driveSearchMaxChatIDs = 20
driveSearchMaxSharerIDs = 20
)
// DriveSearch searches docs/wikis via the v2 doc_wiki/search API using flat
// flags instead of a nested JSON filter, which is friendlier for AI agents and
// `--help` readers.
var DriveSearch = common.Shortcut{
Service: "drive",
Command: "+search",
Description: "Search Lark docs, Wiki, and spreadsheet files with flat filters (Search v2: doc_wiki/search)",
Risk: "read",
Scopes: []string{"search:docs:read"},
AuthTypes: []string{"user"},
HasFormat: true,
Flags: []common.Flag{
{Name: "query", Desc: "search keyword (may be empty to browse by filter only)"},
{Name: "mine", Type: "bool", Desc: "restrict to docs I created (uses current user's open_id)"},
{Name: "creator-ids", Desc: "comma-separated creator open_ids; mutually exclusive with --mine"},
{Name: "edited-since", Desc: "start of [my edited] time window (e.g. 7d, 1m, 1y, 2026-04-01, RFC3339, unix seconds)"},
{Name: "edited-until", Desc: "end of [my edited] time window"},
{Name: "commented-since", Desc: "start of [my commented] time window"},
{Name: "commented-until", Desc: "end of [my commented] time window"},
{Name: "opened-since", Desc: "start of [my opened] time window"},
{Name: "opened-until", Desc: "end of [my opened] time window"},
{Name: "created-since", Desc: "start of [document created] time window"},
{Name: "created-until", Desc: "end of [document created] time window"},
{Name: "doc-types", Desc: "comma-separated types: doc,sheet,bitable,mindnote,file,wiki,docx,folder,catalog,slides,shortcut"},
{Name: "folder-tokens", Desc: "comma-separated folder tokens (doc-only; mutually exclusive with --space-ids)"},
{Name: "space-ids", Desc: "comma-separated wiki space IDs (wiki-only; mutually exclusive with --folder-tokens)"},
{Name: "chat-ids", Desc: "comma-separated chat IDs"},
{Name: "sharer-ids", Desc: "comma-separated sharer open_ids"},
{Name: "only-title", Type: "bool", Desc: "match titles only"},
{Name: "only-comment", Type: "bool", Desc: "search comments only"},
{Name: "sort", Desc: "sort type", Enum: driveSearchSortValues},
{Name: "page-token", Desc: "pagination token from a previous response"},
{Name: "page-size", Default: "15", Desc: "page size (1-20, default 15)"},
},
Validate: func(ctx context.Context, runtime *common.RuntimeContext) error {
return validateDriveSearchIDs(readDriveSearchSpec(runtime))
},
Tips: []string{
"Time flags accept relative (e.g. 7d, 1m, 1y), absolute (2026-04-01, RFC3339), or unix seconds.",
"my_edit_time and my_comment_time are hour-aggregated server-side; sub-hour inputs are snapped and a notice is printed to stderr.",
"Use --mine for a quick \"docs I created\" filter. For other people, use --creator-ids ou_xxx,ou_yyy.",
"--folder-tokens limits to doc-only search; --space-ids limits to wiki-only. They cannot be combined.",
},
DryRun: func(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
spec := readDriveSearchSpec(runtime)
reqBody, notices, err := buildDriveSearchRequest(spec, runtime.UserOpenId(), time.Now())
if err != nil {
return common.NewDryRunAPI().Set("error", err.Error())
}
for _, n := range notices {
fmt.Fprintln(runtime.IO().ErrOut, n)
}
return common.NewDryRunAPI().
POST("/open-apis/search/v2/doc_wiki/search").
Body(reqBody)
},
Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
spec := readDriveSearchSpec(runtime)
reqBody, notices, err := buildDriveSearchRequest(spec, runtime.UserOpenId(), time.Now())
if err != nil {
return err
}
for _, n := range notices {
fmt.Fprintln(runtime.IO().ErrOut, n)
}
data, err := callDriveSearchAPI(runtime, reqBody)
if err != nil {
return err
}
items, _ := data["res_units"].([]interface{})
normalizedItems := addDriveSearchIsoTimeFields(items)
resultData := map[string]interface{}{
"total": data["total"],
"has_more": data["has_more"],
"page_token": data["page_token"],
"results": normalizedItems,
}
runtime.OutFormat(resultData, &output.Meta{Count: len(normalizedItems)}, func(w io.Writer) {
renderDriveSearchTable(w, data, normalizedItems)
})
return nil
},
}
// driveSearchSpec is the parsed flag set for a single +search invocation.
type driveSearchSpec struct {
Query string
PageToken string
PageSize string
Mine bool
CreatorIDs []string
EditedSince string
EditedUntil string
CommentedSince string
CommentedUntil string
OpenedSince string
OpenedUntil string
CreatedSince string
CreatedUntil string
DocTypes []string
FolderTokens []string
SpaceIDs []string
ChatIDs []string
SharerIDs []string
OnlyTitle bool
OnlyComment bool
Sort string
}
func readDriveSearchSpec(runtime *common.RuntimeContext) driveSearchSpec {
return driveSearchSpec{
Query: runtime.Str("query"),
PageToken: runtime.Str("page-token"),
PageSize: runtime.Str("page-size"),
Mine: runtime.Bool("mine"),
CreatorIDs: common.SplitCSV(runtime.Str("creator-ids")),
EditedSince: runtime.Str("edited-since"),
EditedUntil: runtime.Str("edited-until"),
CommentedSince: runtime.Str("commented-since"),
CommentedUntil: runtime.Str("commented-until"),
OpenedSince: runtime.Str("opened-since"),
OpenedUntil: runtime.Str("opened-until"),
CreatedSince: runtime.Str("created-since"),
CreatedUntil: runtime.Str("created-until"),
DocTypes: upperAll(common.SplitCSV(runtime.Str("doc-types"))),
FolderTokens: common.SplitCSV(runtime.Str("folder-tokens")),
SpaceIDs: common.SplitCSV(runtime.Str("space-ids")),
ChatIDs: common.SplitCSV(runtime.Str("chat-ids")),
SharerIDs: common.SplitCSV(runtime.Str("sharer-ids")),
OnlyTitle: runtime.Bool("only-title"),
OnlyComment: runtime.Bool("only-comment"),
Sort: strings.TrimSpace(runtime.Str("sort")),
}
}
// buildDriveSearchRequest turns the parsed spec into the API request body and a
// list of stderr notices (e.g. hour-snap adjustments). It does all validation
// that depends on the combination of flag values.
func buildDriveSearchRequest(spec driveSearchSpec, userOpenID string, now time.Time) (map[string]interface{}, []string, error) {
if spec.Mine && len(spec.CreatorIDs) > 0 {
return nil, nil, output.ErrValidation("cannot combine --mine and --creator-ids")
}
if len(spec.FolderTokens) > 0 && len(spec.SpaceIDs) > 0 {
return nil, nil, output.ErrValidation("cannot combine --folder-tokens and --space-ids; doc and wiki scoped search cannot be combined")
}
if spec.Mine && userOpenID == "" {
return nil, nil, output.ErrValidation("--mine requires a logged-in user open_id, but none is configured; run `lark-cli auth login` or set user open_id in config")
}
if err := validateDocTypes(spec.DocTypes); err != nil {
return nil, nil, err
}
pageSize, err := parseDriveSearchPageSize(spec.PageSize)
if err != nil {
return nil, nil, err
}
request := map[string]interface{}{
"query": spec.Query,
"page_size": pageSize,
}
if spec.PageToken != "" {
request["page_token"] = spec.PageToken
}
filter := map[string]interface{}{}
var notices []string
// open_time is capped at 3 months server-side; if the user's window is
// longer, narrow this request and emit a notice with the remaining slices.
if n, err := clampOpenedTimeWindow(&spec, now); err != nil {
return nil, nil, err
} else if n != "" {
notices = append(notices, n)
}
// Creator identity.
switch {
case spec.Mine:
filter["creator_ids"] = []string{userOpenID}
case len(spec.CreatorIDs) > 0:
filter["creator_ids"] = spec.CreatorIDs
}
// Time dimensions — each fills at most one filter key; hour-aggregated ones
// also contribute notices.
timeDims := []struct {
key string
since, til string
}{
{"my_edit_time", spec.EditedSince, spec.EditedUntil},
{"my_comment_time", spec.CommentedSince, spec.CommentedUntil},
{"open_time", spec.OpenedSince, spec.OpenedUntil},
{"create_time", spec.CreatedSince, spec.CreatedUntil},
}
for _, d := range timeDims {
rng, dimNotices, err := buildTimeRangeFilter(d.key, d.since, d.til, now)
if err != nil {
return nil, nil, err
}
if rng != nil {
filter[d.key] = rng
}
notices = append(notices, dimNotices...)
}
// Scalar scope filters.
if len(spec.DocTypes) > 0 {
filter["doc_types"] = spec.DocTypes
}
if len(spec.ChatIDs) > 0 {
filter["chat_ids"] = spec.ChatIDs
}
if len(spec.SharerIDs) > 0 {
filter["sharer_ids"] = spec.SharerIDs
}
if spec.OnlyTitle {
filter["only_title"] = true
}
if spec.OnlyComment {
filter["only_comment"] = true
}
if spec.Sort != "" {
// Server enum uses "DEFAULT_TYPE" for the default sort; every other
// value upper-cases 1:1.
sortType := strings.ToUpper(spec.Sort)
if sortType == "DEFAULT" {
sortType = "DEFAULT_TYPE"
}
filter["sort_type"] = sortType
}
// Wiki-/folder-scoped variants: keep the shared filter, then add the
// scope-specific key only into the correct side.
switch {
case len(spec.FolderTokens) > 0:
docFilter := cloneDriveSearchFilter(filter)
docFilter["folder_tokens"] = spec.FolderTokens
request["doc_filter"] = docFilter
case len(spec.SpaceIDs) > 0:
wikiFilter := cloneDriveSearchFilter(filter)
wikiFilter["space_ids"] = spec.SpaceIDs
request["wiki_filter"] = wikiFilter
default:
request["doc_filter"] = cloneDriveSearchFilter(filter)
request["wiki_filter"] = cloneDriveSearchFilter(filter)
}
return request, notices, nil
}
func parseDriveSearchPageSize(raw string) (int, error) {
if raw == "" {
return 15, nil
}
n, err := strconv.Atoi(raw)
if err != nil {
return 0, output.ErrValidation("--page-size must be a number, got %q", raw)
}
if n <= 0 {
return 15, nil
}
if n > 20 {
n = 20
}
return n, nil
}
// validateDriveSearchIDs checks open_id / chat_id format and enforces the
// 20-entry cap on chat_ids / sharer_ids before we build the API request,
// so misuse surfaces as a named-flag validation error rather than an opaque
// server-side failure or empty result.
func validateDriveSearchIDs(spec driveSearchSpec) error {
for _, id := range spec.CreatorIDs {
if _, err := common.ValidateUserID(id); err != nil {
return output.ErrValidation("--creator-ids %q: %s", id, err)
}
}
if n := len(spec.ChatIDs); n > driveSearchMaxChatIDs {
return output.ErrValidation("--chat-ids: max %d values per request, got %d", driveSearchMaxChatIDs, n)
}
for _, id := range spec.ChatIDs {
if _, err := common.ValidateChatID(id); err != nil {
return output.ErrValidation("--chat-ids %q: %s", id, err)
}
}
if n := len(spec.SharerIDs); n > driveSearchMaxSharerIDs {
return output.ErrValidation("--sharer-ids: max %d values per request, got %d", driveSearchMaxSharerIDs, n)
}
for _, id := range spec.SharerIDs {
if _, err := common.ValidateUserID(id); err != nil {
return output.ErrValidation("--sharer-ids %q: %s", id, err)
}
}
return nil
}
func validateDocTypes(values []string) error {
for _, v := range values {
// values are already upper-cased by readDriveSearchSpec; compare as-is
// so the filter we emit to the server matches what we validated.
if _, ok := driveSearchDocTypeSet[v]; !ok {
return output.ErrValidation("--doc-types contains unknown value %q (allowed: doc,sheet,bitable,mindnote,file,wiki,docx,folder,catalog,slides,shortcut)", v)
}
}
return nil
}
// upperAll returns a copy of s with every element upper-cased.
func upperAll(s []string) []string {
if len(s) == 0 {
return s
}
out := make([]string, len(s))
for i, v := range s {
out[i] = strings.ToUpper(v)
}
return out
}
// clampOpenedTimeWindow enforces the server-side 3-month cap on open_time by
// narrowing --opened-since / --opened-until to the most recent slice and
// returning a notice that lists every remaining slice, so the agent can
// re-invoke for older ranges. When no clamping is needed, returns ("", nil).
//
// Rules:
// - no --opened-since: skip (no range filter at all)
// - only --opened-since or both set, span ≤ 90 days: skip
// - span in (90, 365] days: clamp current request; spec is mutated in place
// with RFC3339 values so buildTimeRangeFilter parses round-trip
// - span > 365 days: validation error (prevents runaway slice counts)
func clampOpenedTimeWindow(spec *driveSearchSpec, now time.Time) (string, error) {
if spec.OpenedSince == "" {
return "", nil
}
sinceUnix, err := parseTimeValue(spec.OpenedSince, now)
if err != nil {
return "", output.ErrValidation("invalid --opened-since %q: %s", spec.OpenedSince, err)
}
var untilUnix int64
if spec.OpenedUntil != "" {
untilUnix, err = parseTimeValue(spec.OpenedUntil, now)
if err != nil {
return "", output.ErrValidation("invalid --opened-until %q: %s", spec.OpenedUntil, err)
}
} else {
untilUnix = now.Unix()
}
if untilUnix <= sinceUnix {
// Malformed range; let buildTimeRangeFilter / server surface the error.
return "", nil
}
spanSecs := untilUnix - sinceUnix
sliceSecs := int64(driveSearchSliceDays) * 24 * 3600
if spanSecs <= sliceSecs {
return "", nil
}
maxSecs := int64(driveSearchMaxOpenedSpanDays) * 24 * 3600
if spanSecs > maxSecs {
return "", output.ErrValidation(
"--opened-* window spans %d days, exceeds the %d-day (1-year) maximum; narrow the range or run multiple queries",
spanSecs/86400, driveSearchMaxOpenedSpanDays,
)
}
// Build slices newest-to-oldest; last (oldest) slice may be shorter than 90d.
numSlices := int((spanSecs + sliceSecs - 1) / sliceSecs) // ceil
type sliceSpec struct{ start, end int64 }
slices := make([]sliceSpec, numSlices)
cursor := untilUnix
for i := 0; i < numSlices; i++ {
start := cursor - sliceSecs
if start < sinceUnix {
start = sinceUnix
}
slices[i] = sliceSpec{start: start, end: cursor}
cursor = start
}
fmtTime := func(unix int64) string { return time.Unix(unix, 0).Format(time.RFC3339) }
approxMonths := spanSecs / (30 * 24 * 3600)
var b strings.Builder
fmt.Fprintf(&b, "notice: --opened-* window spans %d days (~%d months), exceeds the server-side 3-month (%d-day) limit.\n",
spanSecs/86400, approxMonths, driveSearchSliceDays)
fmt.Fprintf(&b, " this query was narrowed to the most recent slice; %d slices total:\n", numSlices)
// Every slice — including the current one — prints concrete --opened-since
// / --opened-until values so an agent paginating slice 1 can copy them
// verbatim. Reusing the user's original relative time (e.g. "1y") would
// re-resolve against time.Now() on the next call and silently drift the
// window away from any --page-token issued for this call.
for i, s := range slices {
label := fmt.Sprintf("[slice %d/%d]", i+1, numSlices)
if i == 0 {
label = fmt.Sprintf("[slice %d/%d current]", i+1, numSlices)
}
// %-19s pads to "[slice N/M current]" (19 chars at the 5-slice cap).
fmt.Fprintf(&b, " %-19s --opened-since %s --opened-until %s\n",
label, fmtTime(s.start), fmtTime(s.end))
}
fmt.Fprint(&b, " pagination: paginate within a slice via --page-token using that slice's --opened-since / --opened-until values verbatim (NOT the original relative time like '1y' / '8m' — relative times re-resolve against time.Now() and would mismatch the page_token); switch to the next slice's --opened-* flags only after has_more=false, and do not carry --page-token across slices.")
// Rewrite spec so buildTimeRangeFilter emits the clamped window.
spec.OpenedSince = fmtTime(slices[0].start)
spec.OpenedUntil = fmtTime(slices[0].end)
return b.String(), nil
}
// buildTimeRangeFilter parses since/until for one dimension and applies hour
// snapping for server-aggregated fields. Returns nil range when both inputs
// are empty.
func buildTimeRangeFilter(key, since, until string, now time.Time) (map[string]interface{}, []string, error) {
if since == "" && until == "" {
return nil, nil, nil
}
_, hourAggregated := driveSearchHourAggregatedFields[key]
rng := map[string]interface{}{}
var notices []string
if since != "" {
unix, err := parseTimeValue(since, now)
if err != nil {
return nil, nil, output.ErrValidation("invalid --%s-since %q: %s", timeDimCLIName(key), since, err)
}
if hourAggregated && unix%3600 != 0 {
snapped := floorHour(unix)
notices = append(notices, formatHourSnapNotice(key, "start", unix, snapped))
unix = snapped
}
rng["start"] = unix
}
if until != "" {
unix, err := parseTimeValue(until, now)
if err != nil {
return nil, nil, output.ErrValidation("invalid --%s-until %q: %s", timeDimCLIName(key), until, err)
}
if hourAggregated && unix%3600 != 0 {
snapped := ceilHour(unix)
notices = append(notices, formatHourSnapNotice(key, "end", unix, snapped))
unix = snapped
}
rng["end"] = unix
}
return rng, notices, nil
}
// timeDimCLIName maps a filter key back to the CLI flag prefix, for error
// messages that say "--edited-since" rather than "my_edit_time.start".
func timeDimCLIName(key string) string {
switch key {
case "my_edit_time":
return "edited"
case "my_comment_time":
return "commented"
case "open_time":
return "opened"
case "create_time":
return "created"
}
return key
}
func formatHourSnapNotice(key, side string, before, after int64) string {
return fmt.Sprintf("notice: %s has hour-level granularity server-side; %s %s → %s",
key, side,
time.Unix(before, 0).Format("2006-01-02 15:04:05"),
time.Unix(after, 0).Format("2006-01-02 15:04:05"),
)
}
func floorHour(unix int64) int64 {
return unix - (unix % 3600)
}
func ceilHour(unix int64) int64 {
if unix%3600 == 0 {
return unix
}
return floorHour(unix) + 3600
}
var driveSearchRelativeRe = regexp.MustCompile(`^(\d+)([dmy])$`)
// parseTimeValue accepts relative (7d, 1m=30d, 1y=365d), absolute dates in a
// few common layouts, RFC3339, and raw unix seconds.
func parseTimeValue(input string, now time.Time) (int64, error) {
s := strings.TrimSpace(input)
if s == "" {
return 0, fmt.Errorf("empty value")
}
if m := driveSearchRelativeRe.FindStringSubmatch(s); m != nil {
n, _ := strconv.Atoi(m[1])
var days int
switch m[2] {
case "d":
days = n
case "m":
days = n * 30
case "y":
days = n * 365
}
return now.Add(-time.Duration(days) * 24 * time.Hour).Unix(), nil
}
layouts := []string{
time.RFC3339,
"2006-01-02T15:04:05",
"2006-01-02 15:04:05",
"2006-01-02",
}
for _, layout := range layouts {
if t, err := time.ParseInLocation(layout, s, time.Local); err == nil {
return t.Unix(), nil
}
}
// Digit-only string at the end so "20260423" doesn't get misread as unix.
// Real unix seconds for recent times are 10 digits; be conservative and
// require length >= 10 to avoid matching YYYYMMDD. Mirror unixToISO8601's
// ms-vs-s heuristic: 13-digit / >= 1e12 inputs are epoch-millis and get
// normalized to seconds, otherwise a copy-pasted ms timestamp would
// silently parse as a year-57000 unix and then trip the 1-year cap with
// a misleading message.
if len(s) >= 10 {
if n, err := strconv.ParseInt(s, 10, 64); err == nil {
if n >= 1e12 {
n /= 1000
}
return n, nil
}
}
return 0, fmt.Errorf("expected relative (7d/1m/1y), date (YYYY-MM-DD[ HH:MM:SS]), RFC3339, or unix seconds")
}
func callDriveSearchAPI(runtime *common.RuntimeContext, reqBody map[string]interface{}) (map[string]interface{}, error) {
data, err := runtime.CallAPI("POST", "/open-apis/search/v2/doc_wiki/search", nil, reqBody)
if err != nil {
return nil, enrichDriveSearchError(err)
}
return data, nil
}
// enrichDriveSearchError adds a +search-specific hint for known opaque Lark
// codes; other errors pass through unchanged.
func enrichDriveSearchError(err error) error {
var exitErr *output.ExitError
if !errors.As(err, &exitErr) || exitErr.Detail == nil {
return err
}
if exitErr.Detail.Code != driveSearchErrUserNotVisible {
return err
}
detail := *exitErr.Detail
detail.Hint = "one or more open_ids in --creator-ids / --sharer-ids are outside this app's user-visibility scope (this is the app's contact visibility, not the search:docs:read API scope); ask an admin to grant the app visibility to those users in the developer console, or drop the unreachable open_ids"
return &output.ExitError{
Code: exitErr.Code,
Detail: &detail,
Err: exitErr.Err,
Raw: exitErr.Raw,
}
}
func cloneDriveSearchFilter(src map[string]interface{}) map[string]interface{} {
dst := make(map[string]interface{}, len(src))
for k, v := range src {
dst[k] = v
}
return dst
}
// renderDriveSearchTable mirrors the column layout of doc +search so the pretty
// output is consistent for users switching between the two.
func renderDriveSearchTable(w io.Writer, data map[string]interface{}, items []interface{}) {
if len(items) == 0 {
fmt.Fprintln(w, "No matching results found.")
return
}
htmlTagRe := regexp.MustCompile(`</?hb?>`)
var rows []map[string]interface{}
for _, item := range items {
u, _ := item.(map[string]interface{})
if u == nil {
continue
}
var rawTitle string
if s, ok := u["title_highlighted"].(string); ok && s != "" {
rawTitle = s
} else if s, ok := u["title"].(string); ok {
rawTitle = s
}
title := common.TruncateStr(htmlTagRe.ReplaceAllString(rawTitle, ""), 50)
resultMeta, _ := u["result_meta"].(map[string]interface{})
docTypes := ""
if resultMeta != nil {
docTypes = fmt.Sprintf("%v", resultMeta["doc_types"])
}
entityType := fmt.Sprintf("%v", u["entity_type"])
typeStr := docTypes
if typeStr == "" || typeStr == "<nil>" {
typeStr = entityType
}
var url, editTime string
if resultMeta != nil {
if s, ok := resultMeta["url"].(string); ok {
url = s
}
if s, ok := resultMeta["update_time_iso"].(string); ok {
editTime = s
}
}
if len(url) > 80 {
url = url[:80]
}
rows = append(rows, map[string]interface{}{
"type": typeStr,
"title": title,
"edit_time": editTime,
"url": url,
})
}
output.PrintTable(w, rows)
moreHint := ""
hasMore, _ := data["has_more"].(bool)
if hasMore {
moreHint = " (more available, use --format json to get page_token, then --page-token to paginate)"
}
fmt.Fprintf(w, "\n%d result(s)%s\n", len(rows), moreHint)
}
// addDriveSearchIsoTimeFields recursively annotates every `*_time` numeric
// field with a matching `*_time_iso` RFC3339 string, so clients that parse
// JSON output don't have to convert epoch timestamps themselves.
func addDriveSearchIsoTimeFields(value interface{}) []interface{} {
arr, ok := value.([]interface{})
if !ok {
return nil
}
out := make([]interface{}, len(arr))
for i, item := range arr {
out[i] = addDriveSearchIsoTimeFieldsOne(item)
}
return out
}
func addDriveSearchIsoTimeFieldsOne(value interface{}) interface{} {
switch v := value.(type) {
case []interface{}:
result := make([]interface{}, len(v))
for i, item := range v {
result[i] = addDriveSearchIsoTimeFieldsOne(item)
}
return result
case map[string]interface{}:
out := make(map[string]interface{})
for key, item := range v {
if strings.HasSuffix(key, "_time_iso") {
out[key] = item
continue
}
out[key] = addDriveSearchIsoTimeFieldsOne(item)
if strings.HasSuffix(key, "_time") {
// If the input already carries the matching `_iso` sibling,
// the iso-suffix passthrough branch will copy it; don't race
// against it (map iteration order is non-deterministic).
if _, exists := v[key+"_iso"]; exists {
continue
}
if iso := unixToISO8601(item); iso != "" {
out[key+"_iso"] = iso
}
}
}
return out
default:
return value
}
}
func unixToISO8601(v interface{}) string {
if v == nil {
return ""
}
var num float64
switch val := v.(type) {
case float64:
num = val
case json.Number:
parsed, err := val.Float64()
if err != nil {
return ""
}
num = parsed
case string:
parsed, err := strconv.ParseFloat(val, 64)
if err != nil {
return ""
}
num = parsed
case int64:
num = float64(val)
case int:
num = float64(val)
default:
return ""
}
if math.IsInf(num, 0) || math.IsNaN(num) {
return ""
}
secs := int64(num)
if num >= 1e12 {
secs = secs / 1000
}
return time.Unix(secs, 0).Format(time.RFC3339)
}

View File

@@ -0,0 +1,962 @@
// Copyright (c) 2026 Lark Technologies Pte. Ltd.
// SPDX-License-Identifier: MIT
package drive
import (
"bytes"
"encoding/json"
"errors"
"math"
"reflect"
"strings"
"testing"
"time"
"github.com/larksuite/cli/internal/output"
)
// TestClampOpenedTimeWindow covers the 3-month / 1-year boundary logic that
// narrows --opened-since / --opened-until and generates the multi-slice notice.
func TestClampOpenedTimeWindow(t *testing.T) {
t.Parallel()
// Fixed "now" keeps RFC3339 output stable across runs.
now := time.Date(2026, 4, 24, 16, 0, 0, 0, time.UTC)
day := int64(86400)
t.Run("no opened-since: no clamp, no notice", func(t *testing.T) {
t.Parallel()
spec := driveSearchSpec{OpenedUntil: "2026-04-01"}
notice, err := clampOpenedTimeWindow(&spec, now)
if err != nil || notice != "" {
t.Fatalf("got notice=%q err=%v, want both empty", notice, err)
}
if spec.OpenedSince != "" || spec.OpenedUntil != "2026-04-01" {
t.Fatalf("spec mutated unexpectedly: %+v", spec)
}
})
t.Run("span within 90d: no clamp", func(t *testing.T) {
t.Parallel()
spec := driveSearchSpec{OpenedSince: "30d"}
notice, err := clampOpenedTimeWindow(&spec, now)
if err != nil || notice != "" {
t.Fatalf("got notice=%q err=%v, want both empty", notice, err)
}
if spec.OpenedSince != "30d" {
t.Fatalf("spec.OpenedSince mutated: %q", spec.OpenedSince)
}
})
t.Run("exactly 90 days: no clamp", func(t *testing.T) {
t.Parallel()
since := now.Unix() - 90*day
spec := driveSearchSpec{
OpenedSince: time.Unix(since, 0).UTC().Format(time.RFC3339),
}
notice, err := clampOpenedTimeWindow(&spec, now)
if err != nil || notice != "" {
t.Fatalf("got notice=%q err=%v, want no clamp at boundary", notice, err)
}
})
t.Run("91 days: 2-slice clamp", func(t *testing.T) {
t.Parallel()
since := now.Unix() - 91*day
spec := driveSearchSpec{
OpenedSince: time.Unix(since, 0).UTC().Format(time.RFC3339),
}
notice, err := clampOpenedTimeWindow(&spec, now)
if err != nil {
t.Fatalf("unexpected err: %v", err)
}
if !strings.Contains(notice, "2 slices total") {
t.Fatalf("expected '2 slices total' in notice, got:\n%s", notice)
}
// Each slice line — including slice 1 — must spell out concrete
// --opened-since / --opened-until values so a paginating agent can
// copy them verbatim instead of re-using the user's original
// relative time (which would drift against time.Now()).
for _, label := range []string{"[slice 1/2 current]", "[slice 2/2]"} {
var line string
for _, l := range strings.Split(notice, "\n") {
if strings.Contains(l, label) {
line = l
break
}
}
if line == "" {
t.Fatalf("missing %s line, got:\n%s", label, notice)
}
if !strings.Contains(line, "--opened-since ") || !strings.Contains(line, "--opened-until ") {
t.Fatalf("%s line must spell out both flag values, got: %q\nfull notice:\n%s", label, line, notice)
}
}
// After clamp the request window is exactly the most recent 90 days.
clampedSince, err := parseTimeValue(spec.OpenedSince, now)
if err != nil {
t.Fatalf("rewritten opened-since not parseable: %v", err)
}
clampedUntil, err := parseTimeValue(spec.OpenedUntil, now)
if err != nil {
t.Fatalf("rewritten opened-until not parseable: %v", err)
}
if clampedUntil-clampedSince != 90*day {
t.Fatalf("clamped span = %d days, want 90", (clampedUntil-clampedSince)/day)
}
if clampedUntil != now.Unix() {
t.Fatalf("clamped until should default to now; got %d, want %d", clampedUntil, now.Unix())
}
})
t.Run("8 months: 3-slice clamp with shorter tail", func(t *testing.T) {
t.Parallel()
since := now.Unix() - 240*day // 8m ≈ 240 days
spec := driveSearchSpec{
OpenedSince: time.Unix(since, 0).UTC().Format(time.RFC3339),
}
notice, err := clampOpenedTimeWindow(&spec, now)
if err != nil {
t.Fatalf("unexpected err: %v", err)
}
for _, want := range []string{"3 slices total", "[slice 1/3 current]", "[slice 2/3]", "[slice 3/3]"} {
if !strings.Contains(notice, want) {
t.Fatalf("missing %q in notice:\n%s", want, notice)
}
}
})
t.Run("365 days: 5-slice clamp at upper bound", func(t *testing.T) {
t.Parallel()
since := now.Unix() - 365*day
spec := driveSearchSpec{
OpenedSince: time.Unix(since, 0).UTC().Format(time.RFC3339),
}
notice, err := clampOpenedTimeWindow(&spec, now)
if err != nil {
t.Fatalf("365 days should clamp, got err: %v", err)
}
if !strings.Contains(notice, "5 slices total") {
t.Fatalf("expected '5 slices total' for 365-day span, got:\n%s", notice)
}
})
t.Run("over 365 days: hard-cap error", func(t *testing.T) {
t.Parallel()
since := now.Unix() - 366*day
spec := driveSearchSpec{
OpenedSince: time.Unix(since, 0).UTC().Format(time.RFC3339),
}
_, err := clampOpenedTimeWindow(&spec, now)
if err == nil {
t.Fatal("expected error for 366-day span, got nil")
}
if !strings.Contains(err.Error(), "365-day") {
t.Fatalf("error should mention 365-day cap, got: %v", err)
}
})
t.Run("since > until: no clamp, defer to downstream", func(t *testing.T) {
t.Parallel()
spec := driveSearchSpec{
OpenedSince: "2026-04-01",
OpenedUntil: "2026-03-01",
}
notice, err := clampOpenedTimeWindow(&spec, now)
if err != nil || notice != "" {
t.Fatalf("got notice=%q err=%v, want both empty for inverted range", notice, err)
}
})
t.Run("invalid opened-since: validation error", func(t *testing.T) {
t.Parallel()
spec := driveSearchSpec{OpenedSince: "not-a-date"}
_, err := clampOpenedTimeWindow(&spec, now)
if err == nil {
t.Fatal("expected validation error for unparseable since")
}
if !strings.Contains(err.Error(), "--opened-since") {
t.Fatalf("error should name the flag, got: %v", err)
}
})
}
func TestParseDriveSearchPageSize(t *testing.T) {
t.Parallel()
tests := []struct {
name string
raw string
want int
wantErr bool
}{
{"empty defaults to 15", "", 15, false},
{"valid in-range", "10", 10, false},
{"zero falls back to 15", "0", 15, false},
{"negative falls back to 15", "-5", 15, false},
{"clamps to 20 when exceeded", "100", 20, false},
{"non-numeric is a hard error", "abc", 0, true},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
t.Parallel()
got, err := parseDriveSearchPageSize(tt.raw)
if (err != nil) != tt.wantErr {
t.Fatalf("err=%v, wantErr=%v", err, tt.wantErr)
}
if !tt.wantErr && got != tt.want {
t.Fatalf("got %d, want %d", got, tt.want)
}
})
}
}
func TestValidateDocTypes(t *testing.T) {
t.Parallel()
if err := validateDocTypes(nil); err != nil {
t.Fatalf("nil slice should be valid, got: %v", err)
}
if err := validateDocTypes([]string{"DOC", "SHEET", "BITABLE"}); err != nil {
t.Fatalf("known values should pass, got: %v", err)
}
err := validateDocTypes([]string{"DOC", "PIE"})
if err == nil || !strings.Contains(err.Error(), "PIE") {
t.Fatalf("expected error naming the unknown value, got: %v", err)
}
}
func TestUpperAll(t *testing.T) {
t.Parallel()
if got := upperAll(nil); got != nil {
t.Fatalf("nil input should return nil, got %v", got)
}
got := upperAll([]string{"docx", "Sheet", "BITABLE"})
want := []string{"DOCX", "SHEET", "BITABLE"}
if !reflect.DeepEqual(got, want) {
t.Fatalf("got %v, want %v", got, want)
}
}
func TestValidateDriveSearchIDs(t *testing.T) {
t.Parallel()
t.Run("all valid", func(t *testing.T) {
t.Parallel()
spec := driveSearchSpec{
CreatorIDs: []string{"ou_aaa"},
ChatIDs: []string{"oc_xxx"},
SharerIDs: []string{"ou_bbb"},
}
if err := validateDriveSearchIDs(spec); err != nil {
t.Fatalf("unexpected error: %v", err)
}
})
t.Run("bad creator id format", func(t *testing.T) {
t.Parallel()
err := validateDriveSearchIDs(driveSearchSpec{CreatorIDs: []string{"u_bad"}})
if err == nil || !strings.Contains(err.Error(), "--creator-ids") {
t.Fatalf("expected --creator-ids error, got: %v", err)
}
})
t.Run("bad chat id format", func(t *testing.T) {
t.Parallel()
err := validateDriveSearchIDs(driveSearchSpec{ChatIDs: []string{"chat_bad"}})
if err == nil || !strings.Contains(err.Error(), "--chat-ids") {
t.Fatalf("expected --chat-ids error, got: %v", err)
}
})
t.Run("bad sharer id format", func(t *testing.T) {
t.Parallel()
err := validateDriveSearchIDs(driveSearchSpec{SharerIDs: []string{"u_bad"}})
if err == nil || !strings.Contains(err.Error(), "--sharer-ids") {
t.Fatalf("expected --sharer-ids error, got: %v", err)
}
})
t.Run("chat ids exactly at cap is allowed", func(t *testing.T) {
t.Parallel()
ids := make([]string, driveSearchMaxChatIDs)
for i := range ids {
ids[i] = "oc_x"
}
if err := validateDriveSearchIDs(driveSearchSpec{ChatIDs: ids}); err != nil {
t.Fatalf("exactly cap should pass, got: %v", err)
}
})
t.Run("chat ids over cap", func(t *testing.T) {
t.Parallel()
ids := make([]string, driveSearchMaxChatIDs+1)
for i := range ids {
ids[i] = "oc_x"
}
err := validateDriveSearchIDs(driveSearchSpec{ChatIDs: ids})
if err == nil || !strings.Contains(err.Error(), "max") {
t.Fatalf("expected cap error, got: %v", err)
}
})
t.Run("sharer ids exactly at cap is allowed", func(t *testing.T) {
t.Parallel()
ids := make([]string, driveSearchMaxSharerIDs)
for i := range ids {
ids[i] = "ou_x"
}
if err := validateDriveSearchIDs(driveSearchSpec{SharerIDs: ids}); err != nil {
t.Fatalf("exactly cap should pass, got: %v", err)
}
})
t.Run("sharer ids over cap", func(t *testing.T) {
t.Parallel()
ids := make([]string, driveSearchMaxSharerIDs+1)
for i := range ids {
ids[i] = "ou_x"
}
err := validateDriveSearchIDs(driveSearchSpec{SharerIDs: ids})
if err == nil || !strings.Contains(err.Error(), "max") {
t.Fatalf("expected cap error, got: %v", err)
}
})
}
func TestBuildTimeRangeFilter(t *testing.T) {
t.Parallel()
now := time.Date(2026, 4, 24, 16, 0, 0, 0, time.UTC)
t.Run("both empty: nil range, no notice", func(t *testing.T) {
t.Parallel()
rng, notices, err := buildTimeRangeFilter("open_time", "", "", now)
if err != nil || rng != nil || len(notices) != 0 {
t.Fatalf("got rng=%v notices=%v err=%v", rng, notices, err)
}
})
t.Run("open_time passes through without snap", func(t *testing.T) {
t.Parallel()
rng, notices, err := buildTimeRangeFilter("open_time",
"2026-04-20T10:30:45+08:00", "2026-04-21T11:45:30+08:00", now)
if err != nil {
t.Fatalf("err: %v", err)
}
if len(notices) != 0 {
t.Fatalf("open_time should not snap, got notices: %v", notices)
}
if rng["start"] == nil || rng["end"] == nil {
t.Fatalf("range missing endpoints: %v", rng)
}
})
t.Run("my_edit_time snaps sub-hour values", func(t *testing.T) {
t.Parallel()
rng, notices, err := buildTimeRangeFilter("my_edit_time",
"2026-04-20T10:30:45+08:00", "2026-04-21T11:45:30+08:00", now)
if err != nil {
t.Fatalf("err: %v", err)
}
if len(notices) != 2 {
t.Fatalf("expected 2 snap notices (start + end), got %d: %v", len(notices), notices)
}
startUnix := rng["start"].(int64)
endUnix := rng["end"].(int64)
if startUnix%3600 != 0 || endUnix%3600 != 0 {
t.Fatalf("snapped values should align to hour: start=%d end=%d", startUnix, endUnix)
}
})
t.Run("invalid since surfaces with flag name", func(t *testing.T) {
t.Parallel()
_, _, err := buildTimeRangeFilter("my_edit_time", "garbage", "", now)
if err == nil || !strings.Contains(err.Error(), "--edited-since") {
t.Fatalf("expected --edited-since in error, got: %v", err)
}
})
t.Run("invalid until surfaces with flag name", func(t *testing.T) {
t.Parallel()
_, _, err := buildTimeRangeFilter("open_time", "", "garbage", now)
if err == nil || !strings.Contains(err.Error(), "--opened-until") {
t.Fatalf("expected --opened-until in error, got: %v", err)
}
})
}
func TestFloorAndCeilHour(t *testing.T) {
t.Parallel()
// 16:23:45 = unix 1745195025 (arbitrary)
t.Run("floor truncates", func(t *testing.T) {
t.Parallel()
if got := floorHour(1745195025); got%3600 != 0 || got >= 1745195025 {
t.Fatalf("floor(1745195025)=%d invalid", got)
}
})
t.Run("ceil rounds up", func(t *testing.T) {
t.Parallel()
got := ceilHour(1745195025)
if got%3600 != 0 || got <= 1745195025 {
t.Fatalf("ceil(1745195025)=%d invalid", got)
}
})
t.Run("ceil at exact hour is no-op", func(t *testing.T) {
t.Parallel()
exact := int64(1745193600)
if got := ceilHour(exact); got != exact {
t.Fatalf("ceil at hour boundary should be identity, got %d", got)
}
})
}
func TestParseTimeValue(t *testing.T) {
t.Parallel()
now := time.Date(2026, 4, 24, 16, 0, 0, 0, time.Local)
tests := []struct {
name string
input string
wantErr bool
}{
{"empty errors", "", true},
{"7d relative", "7d", false},
{"1m relative", "1m", false},
{"1y relative", "1y", false},
{"date-only YYYY-MM-DD", "2026-04-01", false},
{"datetime with space", "2026-04-01 10:00:00", false},
{"datetime with T", "2026-04-01T10:00:00", false},
{"RFC3339 with offset", "2026-04-01T10:00:00+08:00", false},
{"unix seconds", "1745193600", false},
{"too short to be unix, garbage", "12345", true},
{"YYYYMMDD digits not unix", "20260423", true},
{"unparseable text", "not-a-date", true},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
t.Parallel()
_, err := parseTimeValue(tt.input, now)
if (err != nil) != tt.wantErr {
t.Fatalf("err=%v, wantErr=%v", err, tt.wantErr)
}
})
}
// Sanity: relative units must scale correctly. A regression where "1m"
// silently meant "1 minute" instead of "30 days" would slip past the
// wantErr-only table above; this guards the unit semantics.
t.Run("relative units scale: 7d < 1m < 1y", func(t *testing.T) {
t.Parallel()
got7d, err := parseTimeValue("7d", now)
if err != nil {
t.Fatalf("7d: %v", err)
}
got1m, err := parseTimeValue("1m", now)
if err != nil {
t.Fatalf("1m: %v", err)
}
got1y, err := parseTimeValue("1y", now)
if err != nil {
t.Fatalf("1y: %v", err)
}
// All three are "now minus N days"; larger N means smaller (older) unix.
if !(got1y < got1m && got1m < got7d && got7d < now.Unix()) {
t.Fatalf("expected got1y < got1m < got7d < now; got %d %d %d (now=%d)",
got1y, got1m, got7d, now.Unix())
}
// Spot-check the conversions: "1m" = 30d, "1y" = 365d.
const day = int64(86400)
if now.Unix()-got1m != 30*day {
t.Fatalf("'1m' should resolve to now-30d, got delta %d days", (now.Unix()-got1m)/day)
}
if now.Unix()-got1y != 365*day {
t.Fatalf("'1y' should resolve to now-365d, got delta %d days", (now.Unix()-got1y)/day)
}
})
// Sanity: unix-seconds round-trips exactly (no parsing as date).
t.Run("unix-seconds input round-trips", func(t *testing.T) {
t.Parallel()
got, err := parseTimeValue("1745193600", now)
if err != nil {
t.Fatalf("err: %v", err)
}
if got != 1745193600 {
t.Fatalf("unix round-trip got %d, want 1745193600", got)
}
})
// Regression: a 13-digit epoch-millis timestamp must be normalized to
// seconds. Previously it silently parsed as year-57000 and tripped the
// 1-year cap downstream with a misleading "exceeds 365 days" message.
t.Run("epoch-millis input normalizes to seconds", func(t *testing.T) {
t.Parallel()
got, err := parseTimeValue("1745193600000", now)
if err != nil {
t.Fatalf("err: %v", err)
}
if got != 1745193600 {
t.Fatalf("ms timestamp should normalize to %d seconds, got %d", int64(1745193600), got)
}
})
}
func TestUnixToISO8601(t *testing.T) {
t.Parallel()
const sec int64 = 1745193600 // 2025-04-21 00:00 UTC; only the YYYY-MM-DD prefix is checked below to stay timezone-agnostic
wantPrefix := time.Unix(sec, 0).Format(time.RFC3339)[:10] // YYYY-MM-DD prefix is timezone-stable
tests := []struct {
name string
in interface{}
want string // empty means expect empty result
}{
{"int64", sec, wantPrefix},
{"int", int(sec), wantPrefix},
{"float64", float64(sec), wantPrefix},
{"json.Number", json.Number("1745193600"), wantPrefix},
{"string numeric", "1745193600", wantPrefix},
{"milliseconds get divided", sec * 1000, wantPrefix},
{"nil returns empty", nil, ""},
{"bool ignored", true, ""},
{"unparseable string", "abc", ""},
{"NaN returns empty", math.NaN(), ""},
{"Inf returns empty", math.Inf(1), ""},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
t.Parallel()
got := unixToISO8601(tt.in)
if tt.want == "" {
if got != "" {
t.Fatalf("want empty, got %q", got)
}
return
}
if !strings.HasPrefix(got, tt.want) {
t.Fatalf("got %q, want prefix %q", got, tt.want)
}
})
}
}
func TestAddDriveSearchIsoTimeFields(t *testing.T) {
t.Parallel()
t.Run("non-array input returns nil", func(t *testing.T) {
t.Parallel()
if got := addDriveSearchIsoTimeFields("not-an-array"); got != nil {
t.Fatalf("expected nil, got %v", got)
}
})
t.Run("annotates *_time at top level", func(t *testing.T) {
t.Parallel()
items := []interface{}{
map[string]interface{}{"open_time": int64(1745193600)},
}
row := addDriveSearchIsoTimeFields(items)[0].(map[string]interface{})
if _, ok := row["open_time_iso"].(string); !ok {
t.Fatalf("open_time_iso should have been added, got: %v", row)
}
})
t.Run("recurses into nested map and annotates", func(t *testing.T) {
t.Parallel()
items := []interface{}{
map[string]interface{}{
"result_meta": map[string]interface{}{
"update_time": json.Number("1745193600"),
},
},
}
row := addDriveSearchIsoTimeFields(items)[0].(map[string]interface{})
meta := row["result_meta"].(map[string]interface{})
if _, ok := meta["update_time_iso"].(string); !ok {
t.Fatalf("nested update_time_iso missing, got: %v", meta)
}
})
t.Run("standalone *_time_iso key passes through", func(t *testing.T) {
t.Parallel()
// No sibling *_time key, so the iso-suffix passthrough branch is the
// only one that touches this key — deterministic by construction.
items := []interface{}{
map[string]interface{}{"some_time_iso": "preserved"},
}
row := addDriveSearchIsoTimeFields(items)[0].(map[string]interface{})
if row["some_time_iso"] != "preserved" {
t.Fatalf("existing _time_iso value should pass through, got: %v", row["some_time_iso"])
}
})
// Regression: when both *_time and *_time_iso are present in the same map,
// the pre-existing _iso value must always win, regardless of map iteration
// order. This used to be flaky (a generated iso could overwrite the input
// one depending on which key got visited last).
t.Run("pre-existing *_iso wins over generated when both keys coexist", func(t *testing.T) {
t.Parallel()
const preserved = "PRESERVED-ISO-VALUE"
// Run several times to make a map-iteration-order race surface
// quickly if the guard regresses.
for i := 0; i < 50; i++ {
items := []interface{}{
map[string]interface{}{
"open_time": int64(1745193600),
"open_time_iso": preserved,
},
}
row := addDriveSearchIsoTimeFields(items)[0].(map[string]interface{})
if row["open_time_iso"] != preserved {
t.Fatalf("attempt %d: open_time_iso = %v, want %q (pre-existing must win)",
i, row["open_time_iso"], preserved)
}
}
})
}
func TestEnrichDriveSearchError(t *testing.T) {
t.Parallel()
t.Run("non-ExitError passes through", func(t *testing.T) {
t.Parallel()
orig := errors.New("plain error")
if got := enrichDriveSearchError(orig); got != orig {
t.Fatalf("plain error should pass through unchanged")
}
})
t.Run("ExitError without Detail passes through", func(t *testing.T) {
t.Parallel()
orig := &output.ExitError{Code: 1}
if got := enrichDriveSearchError(orig); got != orig {
t.Fatalf("ExitError without Detail should pass through unchanged")
}
})
t.Run("ExitError with non-matching code passes through", func(t *testing.T) {
t.Parallel()
orig := &output.ExitError{
Code: 1,
Detail: &output.ErrDetail{Code: 12345, Message: "other"},
}
if got := enrichDriveSearchError(orig); got != orig {
t.Fatalf("non-matching code should pass through unchanged")
}
})
t.Run("matching code rewrites Hint without mutating original", func(t *testing.T) {
t.Parallel()
orig := &output.ExitError{
Code: 1,
Detail: &output.ErrDetail{
Code: driveSearchErrUserNotVisible,
Message: "[99992351] user not visible",
Hint: "",
},
}
enriched := enrichDriveSearchError(orig)
eErr, ok := enriched.(*output.ExitError)
if !ok {
t.Fatalf("expected *output.ExitError, got %T", enriched)
}
if eErr == orig {
t.Fatal("should return a new ExitError, not mutate the original")
}
if orig.Detail.Hint != "" {
t.Fatal("original Detail.Hint must remain unchanged")
}
if !strings.Contains(eErr.Detail.Hint, "--creator-ids") {
t.Fatalf("hint should mention --creator-ids, got %q", eErr.Detail.Hint)
}
if eErr.Detail.Message != orig.Detail.Message {
t.Fatalf("Message should be preserved, got %q", eErr.Detail.Message)
}
})
}
func TestCloneDriveSearchFilter(t *testing.T) {
t.Parallel()
src := map[string]interface{}{"a": 1, "b": "x"}
dst := cloneDriveSearchFilter(src)
if !reflect.DeepEqual(src, dst) {
t.Fatalf("clone should equal source")
}
dst["a"] = 99
if src["a"] != 1 {
t.Fatalf("mutating clone should not affect source")
}
}
func TestBuildDriveSearchRequest(t *testing.T) {
t.Parallel()
now := time.Date(2026, 4, 24, 16, 0, 0, 0, time.UTC)
const userOpenID = "ou_self"
t.Run("empty spec emits both filters as empty maps", func(t *testing.T) {
t.Parallel()
req, notices, err := buildDriveSearchRequest(driveSearchSpec{}, userOpenID, now)
if err != nil {
t.Fatalf("err: %v", err)
}
if len(notices) != 0 {
t.Fatalf("expected no notices, got %v", notices)
}
if _, ok := req["doc_filter"].(map[string]interface{}); !ok {
t.Fatalf("doc_filter missing")
}
if _, ok := req["wiki_filter"].(map[string]interface{}); !ok {
t.Fatalf("wiki_filter missing")
}
if req["page_size"] != 15 {
t.Fatalf("default page_size should be 15, got %v", req["page_size"])
}
})
t.Run("--mine fills creator_ids from userOpenID", func(t *testing.T) {
t.Parallel()
req, _, err := buildDriveSearchRequest(driveSearchSpec{Mine: true}, userOpenID, now)
if err != nil {
t.Fatalf("err: %v", err)
}
got := req["doc_filter"].(map[string]interface{})["creator_ids"].([]string)
if len(got) != 1 || got[0] != userOpenID {
t.Fatalf("expected [userOpenID], got %v", got)
}
})
t.Run("--mine without userOpenID errors", func(t *testing.T) {
t.Parallel()
_, _, err := buildDriveSearchRequest(driveSearchSpec{Mine: true}, "", now)
if err == nil || !strings.Contains(err.Error(), "--mine") {
t.Fatalf("expected --mine error, got: %v", err)
}
})
t.Run("--mine + --creator-ids mutually exclusive", func(t *testing.T) {
t.Parallel()
spec := driveSearchSpec{Mine: true, CreatorIDs: []string{"ou_x"}}
_, _, err := buildDriveSearchRequest(spec, userOpenID, now)
if err == nil || !strings.Contains(err.Error(), "--mine") {
t.Fatalf("expected exclusion error, got: %v", err)
}
})
t.Run("--folder-tokens + --space-ids mutually exclusive", func(t *testing.T) {
t.Parallel()
spec := driveSearchSpec{
FolderTokens: []string{"fld_a"},
SpaceIDs: []string{"sp_b"},
}
_, _, err := buildDriveSearchRequest(spec, userOpenID, now)
if err == nil || !strings.Contains(err.Error(), "--folder-tokens") {
t.Fatalf("expected exclusion error, got: %v", err)
}
})
t.Run("--folder-tokens scopes only doc_filter", func(t *testing.T) {
t.Parallel()
spec := driveSearchSpec{FolderTokens: []string{"fld_a"}}
req, _, err := buildDriveSearchRequest(spec, userOpenID, now)
if err != nil {
t.Fatalf("err: %v", err)
}
if _, ok := req["wiki_filter"]; ok {
t.Fatalf("wiki_filter should not be set when --folder-tokens is given")
}
df := req["doc_filter"].(map[string]interface{})
if _, ok := df["folder_tokens"]; !ok {
t.Fatalf("doc_filter must carry folder_tokens")
}
})
t.Run("--space-ids scopes only wiki_filter", func(t *testing.T) {
t.Parallel()
spec := driveSearchSpec{SpaceIDs: []string{"sp_x"}}
req, _, err := buildDriveSearchRequest(spec, userOpenID, now)
if err != nil {
t.Fatalf("err: %v", err)
}
if _, ok := req["doc_filter"]; ok {
t.Fatalf("doc_filter should not be set when --space-ids is given")
}
wf := req["wiki_filter"].(map[string]interface{})
if _, ok := wf["space_ids"]; !ok {
t.Fatalf("wiki_filter must carry space_ids")
}
})
t.Run("sort=default maps to DEFAULT_TYPE", func(t *testing.T) {
t.Parallel()
req, _, err := buildDriveSearchRequest(driveSearchSpec{Sort: "default"}, userOpenID, now)
if err != nil {
t.Fatalf("err: %v", err)
}
if got := req["doc_filter"].(map[string]interface{})["sort_type"]; got != "DEFAULT_TYPE" {
t.Fatalf("sort_type=%v, want DEFAULT_TYPE", got)
}
})
t.Run("sort=edit_time upper-cases 1:1", func(t *testing.T) {
t.Parallel()
req, _, err := buildDriveSearchRequest(driveSearchSpec{Sort: "edit_time"}, userOpenID, now)
if err != nil {
t.Fatalf("err: %v", err)
}
if got := req["doc_filter"].(map[string]interface{})["sort_type"]; got != "EDIT_TIME" {
t.Fatalf("sort_type=%v, want EDIT_TIME", got)
}
})
t.Run("invalid doc-types surfaces", func(t *testing.T) {
t.Parallel()
_, _, err := buildDriveSearchRequest(driveSearchSpec{DocTypes: []string{"PIE"}}, userOpenID, now)
if err == nil || !strings.Contains(err.Error(), "--doc-types") {
t.Fatalf("expected --doc-types error, got: %v", err)
}
})
t.Run("opened-since 8m triggers clamp notice", func(t *testing.T) {
t.Parallel()
spec := driveSearchSpec{
OpenedSince: time.Unix(now.Unix()-240*86400, 0).UTC().Format(time.RFC3339),
}
_, notices, err := buildDriveSearchRequest(spec, userOpenID, now)
if err != nil {
t.Fatalf("err: %v", err)
}
joined := strings.Join(notices, "\n")
if !strings.Contains(joined, "3 slices total") {
t.Fatalf("expected 3-slice clamp notice, got: %s", joined)
}
})
t.Run("scalar filters land in both doc and wiki filters", func(t *testing.T) {
t.Parallel()
spec := driveSearchSpec{
DocTypes: []string{"DOCX"},
ChatIDs: []string{"oc_a"},
OnlyTitle: true,
OnlyComment: true,
}
req, _, err := buildDriveSearchRequest(spec, userOpenID, now)
if err != nil {
t.Fatalf("err: %v", err)
}
df := req["doc_filter"].(map[string]interface{})
wf := req["wiki_filter"].(map[string]interface{})
for _, side := range []map[string]interface{}{df, wf} {
if _, ok := side["doc_types"]; !ok {
t.Fatal("doc_types missing")
}
if _, ok := side["chat_ids"]; !ok {
t.Fatal("chat_ids missing")
}
if side["only_title"] != true {
t.Fatal("only_title missing")
}
if side["only_comment"] != true {
t.Fatal("only_comment missing")
}
}
})
}
func TestRenderDriveSearchTable(t *testing.T) {
t.Parallel()
t.Run("empty items prints fallback message", func(t *testing.T) {
t.Parallel()
var buf bytes.Buffer
renderDriveSearchTable(&buf, map[string]interface{}{}, nil)
if !strings.Contains(buf.String(), "No matching results found") {
t.Fatalf("expected fallback message, got: %s", buf.String())
}
})
t.Run("strips both <h> and <hb> highlight tags", func(t *testing.T) {
t.Parallel()
var buf bytes.Buffer
items := []interface{}{
map[string]interface{}{
"title_highlighted": "<h>hi</h> there <hb>bold</hb>!",
"entity_type": "DOC",
"result_meta": map[string]interface{}{"url": "https://example.com/x"},
},
}
renderDriveSearchTable(&buf, map[string]interface{}{}, items)
out := buf.String()
if strings.Contains(out, "<h>") || strings.Contains(out, "<hb>") || strings.Contains(out, "</h>") || strings.Contains(out, "</hb>") {
t.Fatalf("highlight tags leaked: %s", out)
}
if !strings.Contains(out, "hi there bold!") {
t.Fatalf("plain text should remain after stripping, got: %s", out)
}
})
t.Run("falls back to title when title_highlighted is missing", func(t *testing.T) {
t.Parallel()
var buf bytes.Buffer
items := []interface{}{
map[string]interface{}{
"title": "plain title",
"entity_type": "DOC",
"result_meta": map[string]interface{}{
"url": "https://example.com/x",
"update_time_iso": "2026-04-01T00:00:00Z",
"doc_types": "DOC",
},
},
}
renderDriveSearchTable(&buf, map[string]interface{}{}, items)
out := buf.String()
if !strings.Contains(out, "plain title") {
t.Fatalf("expected fallback title, got: %s", out)
}
if strings.Contains(out, "<nil>") {
t.Fatalf("title fallback should not produce <nil>, got: %s", out)
}
})
// Regression: when result_meta is missing url / update_time_iso (or
// result_meta itself is absent), the table must render empty cells, not
// the literal string "<nil>". This used to leak via fmt.Sprintf("%v",
// nil) before the type-assertion guard was added.
t.Run("missing url and update_time_iso render as empty, not <nil>", func(t *testing.T) {
t.Parallel()
var buf bytes.Buffer
items := []interface{}{
// minimal item: title only, no result_meta keys at all
map[string]interface{}{
"title_highlighted": "row1",
"entity_type": "DOC",
"result_meta": map[string]interface{}{},
},
// item with no result_meta at all
map[string]interface{}{
"title_highlighted": "row2",
"entity_type": "DOC",
},
}
renderDriveSearchTable(&buf, map[string]interface{}{}, items)
out := buf.String()
if strings.Contains(out, "<nil>") {
t.Fatalf("table must not render <nil> for missing url/edit_time, got:\n%s", out)
}
})
t.Run("appends has_more hint when there are more pages", func(t *testing.T) {
t.Parallel()
var buf bytes.Buffer
items := []interface{}{
map[string]interface{}{
"title": "x",
"entity_type": "DOC",
"result_meta": map[string]interface{}{"url": "https://example.com/x"},
},
}
renderDriveSearchTable(&buf, map[string]interface{}{"has_more": true}, items)
if !strings.Contains(buf.String(), "more available") {
t.Fatalf("expected has_more hint, got: %s", buf.String())
}
})
}

View File

@@ -20,5 +20,6 @@ func Shortcuts() []common.Shortcut {
DriveDelete,
DriveTaskResult,
DriveApplyPermission,
DriveSearch,
}
}

View File

@@ -23,6 +23,7 @@ func TestShortcutsIncludesExpectedCommands(t *testing.T) {
"+delete",
"+task_result",
"+apply-permission",
"+search",
}
if len(got) != len(want) {

View File

@@ -37,8 +37,8 @@ lark-cli docs +update --api-version v2 --doc "文档URL或token" --command appen
- 用户说"看一下文档里的图片/附件/素材""预览素材" → 用 `lark-cli docs +media-preview`
- 用户明确说"下载素材" → 用 `lark-cli docs +media-download`
- 如果目标是画板/whiteboard/画板缩略图 → 只能用 `lark-cli docs +media-download --type whiteboard`(不要用 `+media-preview`
- 用户说"找一个表格""按名称搜电子表格""找报表""最近打开的表格" → `lark-cli docs +search` 做资源发现
- `docs +search` 不只搜文档/Wiki结果里会直接返回 `SHEET` 等云空间对象
- 用户说"找一个表格""按名称搜电子表格""找报表""最近打开的表格""最近我编辑过的 xxx"直接`lark-cli drive +search`(参考 [`lark-drive`](../lark-drive/references/lark-drive-search.md))。**老的 `docs +search` 已进入维护期、后续会下线,不要再新增依赖。**
- `drive +search` 结果里会直接返回 `SHEET` / `Base` / `FOLDER` 等云空间对象,是资源发现的统一入口
- 拿到 spreadsheet URL/token 后 → 切到 `lark-sheets` 做对象内部操作
- 用户说"给文档加评论""查看评论""回复评论""给评论加/删除表情 reaction" → 切到 `lark-drive` 处理
- 文档内容中出现嵌入的 `<sheet>``<bitable>``<cite file-type="sheets|bitable">` 标签时 → **必须主动提取 token 并切到对应技能下钻读取内部数据**,不能只呈现标签本身
@@ -51,7 +51,7 @@ lark-cli docs +update --api-version v2 --doc "文档URL或token" --command appen
| `<cite type="doc" file-type="bitable" token="..." table-id="...">` | 同 `<bitable>` | [`lark-base`](../lark-base/SKILL.md) |
| `<synced_reference src-token="..." src-block-id="...">` | `src-token` -> doc_token, `src-block-id` -> block_id | 用 `docs +fetch` 读取 src-token 文档,定位 block |
**补充:** `docs +search` 也承担"先定位云空间对象,再切回对应业务 skill 操作"的资源发现入口角色;当用户口头说"表格/报表"时,也优先从这里开始
**补充:** 云空间资源发现统一走 [`drive +search`](../lark-drive/references/lark-drive-search.md);当用户口头说"表格/报表/最近我编辑过的 xxx"时,也优先从 `drive +search` 开始。老的 `docs +search` 只在沿用 `--filter` JSON 的存量脚本里保留,后续会下线
## Shortcuts推荐优先使用
@@ -59,10 +59,10 @@ Shortcut 是对常用操作的高级封装(`lark-cli docs +<verb> [flags]`
| Shortcut | 说明 |
|----------|------|
| [`+search`](references/lark-doc-search.md) | Search Lark docs, Wiki, and spreadsheet files (Search v2: doc_wiki/search) |
| [`+search`](references/lark-doc-search.md) | ⚠️ **Deprecated — use [`drive +search`](../lark-drive/references/lark-drive-search.md)**. Search Lark docs, Wiki, and spreadsheet files (Search v2: doc_wiki/search). Kept for back-compat; new flows should use the drive-scoped command with flat flags. |
| [`+create`](references/lark-doc-create.md) | Create a Lark document (XML / Markdown) |
| [`+fetch`](references/lark-doc-fetch.md) | Fetch Lark document content (XML / Markdown) |
| [`+update`](references/lark-doc-update.md) | Update a Lark document (str_replace / block_insert_after / block_replace / ...) |
| [`+media-insert`](references/lark-doc-media-insert.md) | Insert a local image or file at the end of a Lark document (4-step orchestration + auto-rollback) |
| [`+media-insert`](references/lark-doc-media-insert.md) | Insert a local image or file at the end of a Lark document (4-step orchestration + auto-rollback). Prefer `--from-clipboard` when the image is already on the system clipboard (screenshots, copy from Feishu/browser); use `--file` only for on-disk sources. |
| [`+media-download`](references/lark-doc-media-download.md) | Download document media or whiteboard thumbnail (auto-detects extension) |
| [`+whiteboard-update`](../lark-whiteboard/references/lark-whiteboard-update.md) | Alias of `whiteboard +update`. Update an existing whiteboard with DSL, Mermaid or PlantUML. Prefer `whiteboard +update`; refer to lark-whiteboard skill for details. |

View File

@@ -1,6 +1,10 @@
# docs +search云空间搜索文档 / Wiki / 电子表格)
> ⚠️ **此命令进入维护期,后续会下线。新用法请使用 [`drive +search`](../../lark-drive/references/lark-drive-search.md)。**
>
> `drive +search` 把所有过滤条件扁平化为独立 flag`--edited-since` / `--mine` / `--doc-types` 等),面向自然语言场景设计,同时新增了 `my_edit_time`(我编辑过)、`my_comment_time`(我评论过)等维度。除非要沿用老脚本里的 `--filter` JSON否则**都应该切到 `drive +search`**。
>
> **前置条件:** 先阅读 [`../lark-shared/SKILL.md`](../../lark-shared/SKILL.md) 了解认证、全局参数和安全规则。
基于 Search v2 接口 `POST /open-apis/search/v2/doc_wiki/search`,以**用户身份**统一搜索云空间对象。

View File

@@ -16,6 +16,7 @@ metadata:
## 快速决策
- 用户要**搜文档 / Wiki / 电子表格 / 多维表格 / 云空间对象**,优先使用 `lark-cli drive +search`。自然语言里"最近我编辑过的"、"我创建的"、"最近一周我打开过的 xxx"、"某人创建的 docx" 等直接映射到扁平 flag避免手写嵌套 JSON。老的 `docs +search` 进入维护期、后续会下线,不要新增对它的依赖。
- 用户要把本地 `.xlsx` / `.csv` / `.base` 导入成 Base / 多维表格 / bitable第一步必须使用 `lark-cli drive +import --type bitable`
- 用户要把本地 `.md` / `.docx` / `.doc` / `.txt` / `.html` 导入成在线文档,使用 `lark-cli drive +import --type docx`
- 用户要把本地 `.xlsx` / `.xls` / `.csv` 导入成电子表格,使用 `lark-cli drive +import --type sheet`
@@ -221,6 +222,7 @@ Shortcut 是对常用操作的高级封装(`lark-cli drive +<verb> [flags]`
| Shortcut | 说明 |
|----------|------|
| [`+search`](references/lark-drive-search.md) | Search Lark docs, Wiki, and spreadsheet files with flat filter flags (preferred over `docs +search`). Natural-language-friendly: `--edited-since`, `--mine`, `--doc-types`, etc. |
| [`+upload`](references/lark-drive-upload.md) | Upload a local file to a Drive folder or wiki node |
| [`+create-folder`](references/lark-drive-create-folder.md) | Create a Drive folder, optionally under a parent folder, with bot auto-grant support |
| [`+download`](references/lark-drive-download.md) | Download a file from Drive to local |

View File

@@ -0,0 +1,239 @@
# drive +search云空间搜索扁平 flag面向自然语言场景
> **前置条件:** 先阅读 [`../lark-shared/SKILL.md`](../../lark-shared/SKILL.md) 了解认证、全局参数和安全规则。
基于 Search v2 接口 `POST /open-apis/search/v2/doc_wiki/search`,以**用户身份**统一搜索云空间对象。
和老的 `docs +search` 相比:
- 把常用过滤条件全部**扁平化为独立 flag**`--edited-since``--mine``--doc-types``--folder-tokens` 等),不再要求用户或 AI 手写嵌套 `--filter` JSON
- 额外暴露了 4 个"我"维度:`my_edit_time`(我编辑过)、`my_comment_time`(我评论过)、`open_time`(我打开过)、`create_time`(文档创建时间)——直接对应用户自然语言里的"最近我编辑过的"、"我评论过的"等表达
- 自动处理 `my_edit_time` / `my_comment_time` 的小时级聚合(服务端存储粒度):亚小时输入会向整点 snap并在 stderr 打出提示
- `--mine` 一键从当前登录用户的 open_id 填 `creator_ids`,不必再先去查 contact
> **资源发现入口统一**`drive +search` 同样返回 `SHEET` / `Base` / `FOLDER` 等全部云空间对象,不只是文档 / Wiki。用户说"找一个表格"、"找报表"、"最近打开的表格"时,也从这里开始;定位后再切到对应业务 skill如 `lark-sheets`)做对象内部操作。
## 命令
> **关键约束:搜索关键词必须通过 `--query` 传递。**
> 正确:`lark-cli drive +search --query "方案"`
> 错误:`lark-cli drive +search 方案`
> `+search` 不接受位置参数;空 `--query` 或省略 `--query` 表示纯靠 filter 浏览(合法)。
### 自然语言 → 命令映射速查
| 用户说 | 命令 |
|---|---|
| 最近一个月我编辑过的文档 | `lark-cli drive +search --query "" --edited-since 1m` |
| 最近一个月我编辑过 且 我评论过的 | `lark-cli drive +search --query "" --edited-since 1m --commented-since 1m` |
| 最近一周我打开过的表格 | `lark-cli drive +search --query "" --opened-since 7d --doc-types sheet` |
| 我创建的所有文档 | `lark-cli drive +search --query "" --mine` |
| 我 30-60 天前创建的文档(粗略"上个月",按 30 天滑窗算) | `lark-cli drive +search --query "" --mine --created-since 2m --created-until 1m` |
| 我 2026 年 3 月创建的文档(精确日历月) | `lark-cli drive +search --query "" --mine --created-since 2026-03-01 --created-until 2026-04-01` |
| 关键词"预算",最近一周我打开过,按编辑时间降序 | `lark-cli drive +search --query 预算 --opened-since 7d --sort edit_time` |
| 某个 wiki space 下、我 30-60 天前创建的 | `lark-cli drive +search --query "" --mine --space-ids space_xxx --created-since 2m --created-until 1m` |
| 张三创建的文档 | `lark-cli drive +search --query "" --creator-ids ou_zhangsan` |
| 我最近 3 个月评论过的 docx | `lark-cli drive +search --query "" --commented-since 3m --doc-types docx` |
### 更多示例
```bash
# 纯关键词搜索
lark-cli drive +search --query "季度总结"
# 使用服务端 query 高级语法(和 docs +search 一致)
lark-cli drive +search --query 'intitle:方案'
lark-cli drive +search --query '"季度 总结"'
lark-cli drive +search --query '方案 OR 草稿'
lark-cli drive +search --query '方案 -草稿'
# 只搜某个文件夹下的文档
lark-cli drive +search --query 方案 --folder-tokens fld_123456
# 只搜某个知识空间下的 Wiki
lark-cli drive +search --query 研发规范 --space-ids space_1234567890fedcba
# 指定群内分享过的文档
lark-cli drive +search --query 方案 --chat-ids oc_1234567890abcdef
# 只搜标题 / 只搜评论
lark-cli drive +search --query 周报 --only-title
lark-cli drive +search --query 延期原因 --only-comment
# 人类可读格式
lark-cli drive +search --query OKR --format pretty
# 翻页(--format json 先拿 page_token
lark-cli drive +search --query 方案 --format json
lark-cli drive +search --query 方案 --page-token '<PAGE_TOKEN>'
```
## 参数
### 核心
| 参数 | 必填 | 说明 |
|---|---|---|
| `--query <text>` | 否 | 搜索关键词;支持服务端高级语法(`intitle:``""``OR``-`)。空字符串或省略表示纯 filter 浏览 |
| `--page-size <n>` | 否 | 每页数量,默认 15最大 20。超过 20 自动 clamp非正数≤0回落 15**非数字值直接返回 validation 错误** |
| `--page-token <token>` | 否 | 上一次响应里的 `page_token`,用于翻页 |
| `--format` | 否 | `json`(默认)/ `pretty` |
### 身份creator 维度)
| 参数 | 映射 | 说明 |
|---|---|---|
| `--mine` | `creator_ids = [当前用户 open_id]` | bool。一键"我创建的";从当前登录用户身份(`runtime.UserOpenId()`)解析 open_id取不到直接报错提示运行 `lark-cli auth login` |
| `--creator-ids ou_x,ou_y` | `creator_ids = [...]` | 显式 open_id 列表,逗号分隔;**与 `--mine` 互斥** |
### 时间维度(每个维度一对 since/until
| 参数 | 映射 API 字段 | 是否小时 snap |
|---|---|---|
| `--edited-since` / `--edited-until` | `my_edit_time.start` / `.end` | ✅ start 向下取整end 向上取整 |
| `--commented-since` / `--commented-until` | `my_comment_time.start` / `.end` | ✅ 同上 |
| `--opened-since` / `--opened-until` | `open_time.start` / `.end` | ❌ 原样透传 |
| `--created-since` / `--created-until` | `create_time.start` / `.end` | ❌ 原样透传(文档创建时间,非"我"语义)|
### 作用域
| 参数 | 映射 | 说明 |
|---|---|---|
| `--doc-types docx,sheet` | `doc_types` | 逗号分隔。允许值:`doc,sheet,bitable,mindnote,file,wiki,docx,folder,catalog,slides,shortcut` |
| `--folder-tokens fld_a,fld_b` | `folder_tokens`(仅 doc_filter | 存在时只发 `doc_filter`**与 `--space-ids` 互斥** |
| `--space-ids sp_x` | `space_ids`(仅 wiki_filter | 存在时只发 `wiki_filter`**与 `--folder-tokens` 互斥** |
| `--chat-ids oc_x` | `chat_ids` | 逗号分隔 |
| `--sharer-ids ou_x` | `sharer_ids` | 逗号分隔open_id |
### 其他
| 参数 | 映射 | 说明 |
|---|---|---|
| `--only-title` | `only_title: true` | bool |
| `--only-comment` | `only_comment: true` | bool |
| `--sort <value>` | `sort_type`(转大写枚举) | 允许值:`default, edit_time, edit_time_asc, open_time, create_time` |
> `--sort`CLI 只暴露服务端**正式支持**的 5 个值。服务端 enum 里 `CREATE_TIME_ASC` 协议标注"暂不支持"`ENTITY_CREATE_TIME_ASC` / `ENTITY_CREATE_TIME_DESC` 已废弃CLI 直接不放出来,传了会被 cobra enum 校验拒掉。
## 时间值格式
所有 `--*-since` / `--*-until` 共用:
| 输入 | 含义 |
|---|---|
| `7d` / `30d` | N 天前的当前时刻 |
| `1m` | 30 天前(固定 30 天,**不是**日历月)|
| `3m` / `6m` | 90 / 180 天前 |
| `1y` | 365 天前 |
| `2026-04-01` | 本地时区 00:00:00 |
| `2026-04-01 10:00:00` / `2026-04-01T10:00:00` | 本地时区具体时刻 |
| `2026-04-01T10:00:00+08:00` | RFC3339 带时区 |
| `1743523200`(≥ 10 位纯数字)| Unix 秒直接透传 |
> `m` 绑定 month30 天),不支持 minute——因为 `my_edit_time` / `my_comment_time` 在服务端是小时聚合,分钟粒度没意义。
## 小时聚合my_edit_time / my_comment_time
服务端对这两个字段按整点聚合,亚小时输入会被 CLI 向整点对齐:
```text
start: floor 到整点 16:23:45 → 16:00:00
end: ceil 到整点 16:23:45 → 17:00:00
```
发生对齐时stderr 会打印一条 notice例如
```text
notice: my_edit_time has hour-level granularity server-side;
start 2026-04-22 16:23:00 → 2026-04-22 16:00:00
end 2026-04-22 16:28:00 → 2026-04-22 17:00:00
```
stdout 的 JSON 输出不受影响。`open_time` / `create_time` 不做 snap。
## 输出
- `--format json`(默认):`{ total, has_more, page_token, results: [...] }`;所有 `*_time` 字段递归补 `*_time_iso`
- `--format pretty`4 列 table —— `type | title | edit_time | url`
- `title_highlighted` / `summary_highlighted` 可能包含 `<h>` / `<hb>` 高亮标签,客户端对比前需先剥离
> **注意**:返回体里的 `total` 字段不够准确(官方确认,仅供参考)。需要精确统计的场景,按实际 `results` 做去重和累加,不要把 `total` 当结果数承诺。
## 决策规则
- **和 `docs +search` 的选择**:优先使用 `drive +search`(本指令),不要再用 `docs +search``docs +search` 进入维护期、后续会下线。
- **身份快捷方式**:只要用户说"我创建的",直接 `--mine` 即可,不需要先查 contact 拿 open_id。
- **时间维度选择**
- "我编辑的"、"我修改的" → `--edited-since` / `--edited-until`
- "我评论的"、"我回复过的" → `--commented-since` / `--commented-until`
- "我看过的"、"我打开过的"、"最近看过的" → `--opened-since` / `--opened-until`
- "创建于"、"新建的"(文档整体维度,与"我"无关)→ `--created-since` / `--created-until`
- **作用域选择**
- "某个文件夹下" → `--folder-tokens`doc-only
- "某个 wiki 空间下" → `--space-ids`wiki-only
- 两者不能同时使用,混用会报错
- **身份 flag 互斥**`--mine``--creator-ids` 不要同时传,会直接报错。"我和张三创建的" 用 `--creator-ids ou_me,ou_zhangsan`(需要先拿到自己 open_id但这种场景少见
- **实体补全**
- 用户说"某个群里",先用 `lark-im``chat_id`
- 用户说"某人创建/分享的"(非自己),先用 `lark-contact` 查 open_id再填 `--creator-ids` / `--sharer-ids`
- **查询语义下推**`--query` 支持的服务端高级语法(`intitle:``""``OR``-`)优先使用,不要先模糊搜再在客户端二次过滤。
- **时间表达**
- 模糊相对时间("最近半年"、"过去 30 天"、"最近一周")→ `--*-since 6m` / `--*-since 30d` / `--*-since 7d`,不展开成 ISO 时间
- **日历表达**"上个月"、"上周"、"本月"、"前年"、"今年 3 月"等明确日历单位)→ **必须算出绝对 `YYYY-MM-DD` 边界**(如"上个月" = 上一个日历月的 1 号 → 当月 1 号),**不要近似成 `1m`/`2m`**CLI 里 `m` 是固定 30 天、`y` 固定 365 天,跟日历差 0-3 天,月末月初尤其容易偏出去
- 绝对日期 → 直接 `YYYY-MM-DD` 或 RFC3339
- **分页策略**:默认只返回第一页,并说明 `has_more` 和下一页命令。只有用户明确要"全部 / 全量 / 继续翻"才继续。单轮翻页上限 5 页。
- **原始返回**:用户要求"原始数据"、"接口返回"时用 `--format json`,不做客户端精确过滤或摘要重写。
## 权限
| 操作 | 所需 scope |
|---|---|
| 搜索云空间对象(文档 / Wiki / 表格等资源发现) | `search:docs:read` |
## 常见错误
| code | 含义 | 处理 |
|---|---|---|
| `99992351` | `--creator-ids` / `--sharer-ids` 里有 open_id 超出**应用的通讯录可见范围**,服务端拒绝识别 | 让管理员在开发者后台把这些用户加进应用的"通讯录可见性"授权里;或把超出范围的 open_id 从参数里去掉。这和 `search:docs:read` scope 不是一回事 —— 是"应用能看见哪些人"而不是"应用能调用哪个接口" |
## 时间范围自动裁剪(`--opened-*` 专有)
服务端对 `open_time` 过滤**每次请求最多支持 3 个月**90 天)窗口。其他三个时间维度(`--edited-*` / `--commented-*` / `--created-*`**不受影响**。
CLI 在发请求前会检查 `--opened-since` 到有效 `--opened-until`(没传则取 `now`)的跨度:
| 跨度 | 行为 |
|---|---|
| ≤ 90 天 | 原样透传 |
| 91 ~ 365 天 | **自动裁剪**到"最近一个 90 天 slice"stderr 打一条 notice 列出所有剩余 slice 的 `--opened-since` / `--opened-until` 参数值 |
| > 365 天 | 直接报 validation 错,要求缩小范围或自行拆分多次查询 |
Notice 示例(用户原本要求"过去 8 个月",会被拆成 3 个 slice
```text
notice: --opened-* window spans 240 days (~8 months), exceeds the server-side 3-month (90-day) limit.
this query was narrowed to the most recent slice; 3 slices total:
[slice 1/3 current] --opened-since 2026-01-24T21:54:02+08:00 --opened-until 2026-04-24T21:54:02+08:00
[slice 2/3] --opened-since 2025-10-26T21:54:02+08:00 --opened-until 2026-01-24T21:54:02+08:00
[slice 3/3] --opened-since 2025-08-27T21:54:02+08:00 --opened-until 2025-10-26T21:54:02+08:00
pagination: paginate within a slice via --page-token using that slice's --opened-since / --opened-until values verbatim (NOT the original relative time like '1y' / '8m' — relative times re-resolve against time.Now() and would mismatch the page_token); switch to the next slice's --opened-* flags only after has_more=false, and do not carry --page-token across slices.
```
### Agent 看到 notice 时的处理
**标准流程(分页 × slice 的先后顺序):**
1. **跑 slice 1**(本次请求已自动裁剪到这个窗口),把结果呈现给用户
2. **先在当前 slice 内翻页**:返回的 `has_more = true` 且用户想看更多时,把 `--opened-since` / `--opened-until` 改成 notice 里 `[slice 1/N current]` 行给出的**具体时间值****不要继续用原始的 `--opened-since 1y` 这种相对值**——CLI 每次调用都按 `time.Now()` 重算窗口,相对值 + `--page-token` 一起跑会让 page_token 绑到一个漂移的窗口上、结果静默失真),加 `--page-token` 继续翻,直到 `has_more = false`
3. **再切换到下一个 slice**:当前 slice 翻完后,如果用户还要"更老的",用 notice 里列的 slice 2 的 `--opened-since` / `--opened-until` 值,**其他 flag`--query``--doc-types``--page-size``--sort`……)保持原样,`--page-token` 不带**,重新发请求
4. **依次递推**slice 2 翻完后切 slice 3以此类推
5. 用户只对最近一段感兴趣时,跳过第 3 步及以后 —— 避免无意义的 API 调用
> `--page-token` 只在单 slice 上下文内有效;切 slice 时不要把上一个 slice 的 `page_token` 带过去。
### 注意事项
- `--sort` 在**单 slice 内部**是正确的。跨 slice 的全局 sort例如"过去一年我打开过的,按 edit_time desc 排")不被 CLI 保证,需要 agent 自行拉完多个 slice 后在客户端 re-sort 再呈现
- 裁剪只改 request 发出去的 `open_time` 范围,`--query` / 其他 filter 不动
- 最后一个最老的slice 常常不足 90 天,这是正常的截断

View File

@@ -0,0 +1,338 @@
// Copyright (c) 2026 Lark Technologies Pte. Ltd.
// SPDX-License-Identifier: MIT
package drive
import (
"context"
"strings"
"testing"
"time"
clie2e "github.com/larksuite/cli/tests/cli_e2e"
"github.com/stretchr/testify/require"
"github.com/tidwall/gjson"
)
// TestDriveSearchDryRun_RequestShape locks in the dry-run request body so
// agents that key off of stdout (URL, doc_filter / wiki_filter, scalar
// filters) don't silently regress. Run end-to-end so cobra flag parsing,
// readDriveSearchSpec, and the dry-run renderer all execute against the
// real binary.
//
// Fake credentials are sufficient because --dry-run short-circuits before
// any network call.
func TestDriveSearchDryRun_RequestShape(t *testing.T) {
setDriveSearchE2EEnv(t)
tests := []struct {
name string
args []string
// JSONPath assertions over the dry-run body.
wantURL string
wantQuery string
wantDocFilter bool
wantWikiFilter bool
wantDocFilterFields map[string]string // gjson path under api.0.body.doc_filter -> string value (or "" to require existence only)
wantWikiFilterFields map[string]string
}{
{
name: "basic --query emits both filters",
args: []string{
"drive", "+search",
"--query", "season report",
"--page-size", "5",
"--dry-run",
},
wantURL: "/open-apis/search/v2/doc_wiki/search",
wantQuery: "season report",
wantDocFilter: true,
wantWikiFilter: true,
},
{
name: "--folder-tokens scopes to doc_filter only",
args: []string{
"drive", "+search",
"--query", "x",
"--folder-tokens", "fld_aaa,fld_bbb",
"--dry-run",
},
wantURL: "/open-apis/search/v2/doc_wiki/search",
wantQuery: "x",
wantDocFilter: true,
wantDocFilterFields: map[string]string{
"folder_tokens.0": "fld_aaa",
"folder_tokens.1": "fld_bbb",
},
},
{
name: "--space-ids scopes to wiki_filter only",
args: []string{
"drive", "+search",
"--query", "x",
"--space-ids", "sp_xxx",
"--dry-run",
},
wantURL: "/open-apis/search/v2/doc_wiki/search",
wantQuery: "x",
wantWikiFilter: true,
wantWikiFilterFields: map[string]string{
"space_ids.0": "sp_xxx",
},
},
{
name: "--sort default maps to DEFAULT_TYPE in body",
args: []string{
"drive", "+search",
"--query", "x",
"--sort", "default",
"--dry-run",
},
wantURL: "/open-apis/search/v2/doc_wiki/search",
wantQuery: "x",
wantDocFilter: true,
wantWikiFilter: true,
wantDocFilterFields: map[string]string{
"sort_type": "DEFAULT_TYPE",
},
},
{
name: "mixed-case --doc-types is normalized to upper case in body",
args: []string{
"drive", "+search",
"--query", "x",
"--doc-types", "docx,Sheet,BITABLE",
"--dry-run",
},
wantURL: "/open-apis/search/v2/doc_wiki/search",
wantQuery: "x",
wantDocFilter: true,
wantWikiFilter: true,
wantDocFilterFields: map[string]string{
"doc_types.0": "DOCX",
"doc_types.1": "SHEET",
"doc_types.2": "BITABLE",
},
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
t.Cleanup(cancel)
result, err := clie2e.RunCmd(ctx, clie2e.Request{
Args: tt.args,
DefaultAs: "user",
})
require.NoError(t, err)
result.AssertExitCode(t, 0)
out := result.Stdout
if got := gjson.Get(out, "api.0.method").String(); got != "POST" {
t.Fatalf("method=%q, want POST\nstdout:\n%s", got, out)
}
if got := gjson.Get(out, "api.0.url").String(); got != tt.wantURL {
t.Fatalf("url=%q, want %q\nstdout:\n%s", got, tt.wantURL, out)
}
if got := gjson.Get(out, "api.0.body.query").String(); got != tt.wantQuery {
t.Fatalf("body.query=%q, want %q\nstdout:\n%s", got, tt.wantQuery, out)
}
if tt.wantDocFilter && !gjson.Get(out, "api.0.body.doc_filter").Exists() {
t.Fatalf("doc_filter missing\nstdout:\n%s", out)
}
if !tt.wantDocFilter && gjson.Get(out, "api.0.body.doc_filter").Exists() {
t.Fatalf("doc_filter should be omitted\nstdout:\n%s", out)
}
if tt.wantWikiFilter && !gjson.Get(out, "api.0.body.wiki_filter").Exists() {
t.Fatalf("wiki_filter missing\nstdout:\n%s", out)
}
if !tt.wantWikiFilter && gjson.Get(out, "api.0.body.wiki_filter").Exists() {
t.Fatalf("wiki_filter should be omitted\nstdout:\n%s", out)
}
for path, want := range tt.wantDocFilterFields {
if got := gjson.Get(out, "api.0.body.doc_filter."+path).String(); got != want {
t.Fatalf("doc_filter.%s=%q, want %q\nstdout:\n%s", path, got, want, out)
}
}
for path, want := range tt.wantWikiFilterFields {
if got := gjson.Get(out, "api.0.body.wiki_filter."+path).String(); got != want {
t.Fatalf("wiki_filter.%s=%q, want %q\nstdout:\n%s", path, got, want, out)
}
}
})
}
}
// TestDriveSearchDryRun_OpenedClamping locks in the agent-facing slice
// notice for --opened-* spans over 90 days: the request body must carry
// the most recent 90-day window, and stderr must list slice N's flag
// values verbatim so the agent can re-invoke for older ranges.
func TestDriveSearchDryRun_OpenedClamping(t *testing.T) {
setDriveSearchE2EEnv(t)
ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
t.Cleanup(cancel)
result, err := clie2e.RunCmd(ctx, clie2e.Request{
Args: []string{
"drive", "+search",
"--query", "x",
"--opened-since", "8m",
"--dry-run",
},
DefaultAs: "user",
})
require.NoError(t, err)
result.AssertExitCode(t, 0)
// Notice goes to stderr alongside other dimension notices.
for _, want := range []string{
"--opened-* window spans",
"3 slices total",
"[slice 1/3 current]",
"[slice 2/3]",
"[slice 3/3]",
"--page-token",
} {
if !strings.Contains(result.Stderr, want) {
t.Fatalf("notice missing %q\nstderr:\n%s", want, result.Stderr)
}
}
// Slice 1 specifically must spell out concrete --opened-* flag values
// (not just the timestamps in arrow form): an agent paginating slice 1
// has to copy these verbatim, otherwise reusing the original relative
// time '8m' would drift the window against time.Now() and mismatch the
// page_token.
for _, label := range []string{"[slice 1/3 current]", "[slice 2/3]", "[slice 3/3]"} {
var line string
for _, l := range strings.Split(result.Stderr, "\n") {
if strings.Contains(l, label) {
line = l
break
}
}
if !strings.Contains(line, "--opened-since ") || !strings.Contains(line, "--opened-until ") {
t.Fatalf("%s line must spell out both flags, got %q\nfull stderr:\n%s", label, line, result.Stderr)
}
}
// And the request body's open_time must reflect the clamped window
// (start and end both present, span = 90 days exactly).
body := result.Stdout
start := gjson.Get(body, "api.0.body.doc_filter.open_time.start").Int()
end := gjson.Get(body, "api.0.body.doc_filter.open_time.end").Int()
if start == 0 || end == 0 {
t.Fatalf("doc_filter.open_time.start/end missing\nstdout:\n%s", body)
}
if delta := end - start; delta != 90*86400 {
t.Fatalf("clamped span = %d seconds, want %d (90 days)\nstdout:\n%s", delta, 90*86400, body)
}
}
// TestDriveSearchDryRun_RejectsOpenedOver1Year locks in the hard cap: a
// --opened-* span beyond 365 days fails validation up front and never
// reaches the API. Important because the alternative (silent slicing into
// many windows) would produce a rate-limit / runaway request loop.
//
// Dry-run captures spec-level validation errors into the JSON envelope's
// `error` field (api list comes back empty); the process still exits 0
// because the dry-run itself succeeded — it just told you what would have
// failed at execution time.
func TestDriveSearchDryRun_RejectsOpenedOver1Year(t *testing.T) {
setDriveSearchE2EEnv(t)
ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
t.Cleanup(cancel)
result, err := clie2e.RunCmd(ctx, clie2e.Request{
Args: []string{
"drive", "+search",
"--query", "x",
"--opened-since", "2y",
"--dry-run",
},
DefaultAs: "user",
})
require.NoError(t, err)
result.AssertExitCode(t, 0)
if api := gjson.Get(result.Stdout, "api"); api.IsArray() && len(api.Array()) > 0 {
t.Fatalf("dry-run api list must be empty when validation fails\nstdout:\n%s", result.Stdout)
}
errMsg := gjson.Get(result.Stdout, "error").String()
if !strings.Contains(errMsg, "365-day") {
t.Fatalf("expected 365-day cap message in dry-run error, got %q\nstdout:\n%s", errMsg, result.Stdout)
}
}
// TestDriveSearchDryRun_RejectsInvalidSort locks in the cobra Enum guard.
// CLI intentionally exposes only 5 sort values (default, edit_time,
// edit_time_asc, open_time, create_time); the deprecated /
// not-supported server enum values must be rejected before reaching the
// request layer.
func TestDriveSearchDryRun_RejectsInvalidSort(t *testing.T) {
setDriveSearchE2EEnv(t)
ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
t.Cleanup(cancel)
result, err := clie2e.RunCmd(ctx, clie2e.Request{
Args: []string{
"drive", "+search",
"--query", "x",
"--sort", "create_time_asc",
"--dry-run",
},
DefaultAs: "user",
})
require.NoError(t, err)
if result.ExitCode == 0 {
t.Fatalf("invalid sort must be rejected, got exit=0\nstdout:\n%s", result.Stdout)
}
combined := result.Stdout + "\n" + result.Stderr
// Pin to the flag name (with dashes) rather than the bare word "sort",
// which would also match "transport" / "sortable" / etc.
if !strings.Contains(combined, "--sort") {
t.Fatalf("expected --sort error message, got:\nstdout:\n%s\nstderr:\n%s", result.Stdout, result.Stderr)
}
}
// TestDriveSearchDryRun_RejectsBadDocType verifies the doc-types validator
// is wired at the dry-run path: an unknown enum value surfaces as a
// validation error inside the dry-run JSON envelope rather than reaching
// the server. The process still exits 0 (see RejectsOpenedOver1Year).
func TestDriveSearchDryRun_RejectsBadDocType(t *testing.T) {
setDriveSearchE2EEnv(t)
ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
t.Cleanup(cancel)
result, err := clie2e.RunCmd(ctx, clie2e.Request{
Args: []string{
"drive", "+search",
"--query", "x",
"--doc-types", "docx,pie",
"--dry-run",
},
DefaultAs: "user",
})
require.NoError(t, err)
result.AssertExitCode(t, 0)
if api := gjson.Get(result.Stdout, "api"); api.IsArray() && len(api.Array()) > 0 {
t.Fatalf("dry-run api list must be empty when validation fails\nstdout:\n%s", result.Stdout)
}
errMsg := gjson.Get(result.Stdout, "error").String()
if !strings.Contains(errMsg, "--doc-types") {
t.Fatalf("expected --doc-types error in dry-run, got %q\nstdout:\n%s", errMsg, result.Stdout)
}
}
func setDriveSearchE2EEnv(t *testing.T) {
t.Helper()
t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
t.Setenv("LARKSUITE_CLI_APP_ID", "drive_search_e2e_app")
t.Setenv("LARKSUITE_CLI_APP_SECRET", "drive_search_e2e_secret")
t.Setenv("LARKSUITE_CLI_BRAND", "feishu")
}