larksuite__cli/shortcuts/slides/slides_create.go

// Copyright (c) 2026 Lark Technologies Pte. Ltd.
// SPDX-License-Identifier: MIT

package slides

import (
	"context"
	"encoding/json"
	"fmt"
	"path/filepath"
	"strings"

	"github.com/larksuite/cli/errs"
	"github.com/larksuite/cli/internal/cmdutil"
	"github.com/larksuite/cli/internal/validate"
	"github.com/larksuite/cli/shortcuts/common"
)

const (
	defaultPresentationWidth  = 960
	defaultPresentationHeight = 540
	maxSlidesPerCreate        = 10
)

// SlidesCreate creates a new Lark Slides presentation with bot auto-grant.
var SlidesCreate = common.Shortcut{
	Service:     "slides",
	Command:     "+create",
	Description: "Create a Lark Slides presentation",
	Risk:        "write",
	AuthTypes:   []string{"user", "bot"},
	// docs:document.media:upload is required by the @-placeholder upload path.
	// Declared up-front (matching the convention used by other multi-API shortcuts
	// like wiki_move) so the pre-flight check fails fast and lark-cli's
	// auth login --scope hint guides the user, instead of leaving an orphaned
	// empty presentation when the in-flight upload 403s.
	// NB: no drive scope here on purpose — slides creation never touches drive;
	// the presentation URL is built locally (see Execute), so we don't gate a
	// drive-free operation behind a drive scope.
	Scopes: []string{"slides:presentation:create", "slides:presentation:write_only", "docs:document.media:upload"},
	Flags: []common.Flag{
		{Name: "title", Desc: "presentation title"},
		// The "(supports @file, - reads stdin ...)" suffix is appended from Input
		// by the framework, so it is not spelled out in Desc.
		{Name: "slides", Desc: "slide content JSON array (each element is a <slide> XML string, max 10; for more pages, create first then add them one at a time with slides +add-slide). <img src=\"@./local.png\"> placeholders are auto-uploaded and replaced with file_token.", Input: []string{common.File, common.Stdin}},
		{Name: "slide", Type: "string_array", Desc: "one complete <slide> XML document, or @path to read one from a file; repeat once per page (max 10) and the CLI assembles the array for you, so no JSON escaping is needed. <img src=\"@./local.png\"> placeholders are handled as with --slides. Mutually exclusive with --slides."},
	},
	Validate: func(ctx context.Context, runtime *common.RuntimeContext) error {
		slides, param, err := createSlideContents(runtime)
		if err != nil {
			return err
		}
		if len(slides) == 0 {
			return nil
		}
		// Validate placeholder paths up front so we don't create a presentation
		// only to fail mid-way on a missing local file.
		return validateImagePlaceholderFiles(runtime, param, extractImagePlaceholderPaths(slides))
	},
	DryRun: func(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
		title := effectiveTitle(runtime.Str("title"))
		slides, _, err := createSlideContents(runtime)
		if err != nil {
			return common.NewDryRunAPI().Set("error", err.Error())
		}
		createBody := map[string]interface{}{
			"xml_presentation": map[string]interface{}{"content": buildPresentationXML(title)},
		}

		dry := common.NewDryRunAPI()

		if len(slides) == 0 {
			dry.Desc("Create empty presentation").
				POST("/open-apis/slides_ai/v1/xml_presentations").
				Body(createBody)
		} else {
			n := len(slides)
			placeholders := extractImagePlaceholderPaths(slides)
			total := n + 1 + len(placeholders)

			descSuffix := ""
			if len(placeholders) > 0 {
				descSuffix = fmt.Sprintf(" + upload %d image(s)", len(placeholders))
			}
			dry.Desc(fmt.Sprintf("Create presentation%s + add %d slide(s)", descSuffix, n)).
				POST("/open-apis/slides_ai/v1/xml_presentations").
				Desc(fmt.Sprintf("[1/%d] Create presentation", total)).
				Body(createBody)

			// Upload steps come right after creation so they can use the new
			// presentation_id as parent_node.
			for i, path := range placeholders {
				appendSlidesUploadDryRun(dry, path, "<xml_presentation_id>", i+2)
			}

			slideStepStart := 2 + len(placeholders)
			slideDescSuffix := ""
			if len(placeholders) > 0 {
				slideDescSuffix = " (img placeholders auto-replaced)"
			}
			for i, slideXML := range slides {
				dry.POST("/open-apis/slides_ai/v1/xml_presentations/<xml_presentation_id>/slide").
					Desc(fmt.Sprintf("[%d/%d] Add slide %d%s", slideStepStart+i, total, i+1, slideDescSuffix)).
					Body(map[string]interface{}{
						"slide": map[string]interface{}{"content": slideXML},
					})
			}
		}

		if runtime.IsBot() {
			dry.Desc("After creation succeeds in bot mode, the CLI will also try to grant the current CLI user full_access on the new presentation.")
		}
		return dry
	},
	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
		title := effectiveTitle(runtime.Str("title"))
		content := buildPresentationXML(title)
		// Resolve @path inputs before the create call: a bad path must not leave
		// an orphaned empty presentation behind.
		slides, param, err := createSlideContents(runtime)
		if err != nil {
			return err
		}

		// Step 1: Create presentation
		data, err := runtime.CallAPITyped(
			"POST",
			"/open-apis/slides_ai/v1/xml_presentations",
			nil,
			map[string]interface{}{
				"xml_presentation": map[string]interface{}{
					"content": content,
				},
			},
		)
		if err != nil {
			return err
		}

		presentationID := common.GetString(data, "xml_presentation_id")
		if presentationID == "" {
			return errs.NewInternalError(errs.SubtypeInvalidResponse, "slides create returned no xml_presentation_id")
		}

		result := map[string]interface{}{
			"xml_presentation_id": presentationID,
			"title":               title,
		}
		if revisionID := common.GetFloat(data, "revision_id"); revisionID > 0 {
			result["revision_id"] = int(revisionID)
		}
		if issues, ok := data["issues"]; ok {
			result["issues"] = issues
		}

		// Step 2: Add slides if provided
		if len(slides) > 0 {
			// Step 1.5: Upload any @path placeholders, then rewrite slide XML
			// with the resulting file_tokens. Uploads run after creation so
			// they can use the new presentation_id as parent_node.
			placeholders := extractImagePlaceholderPaths(slides)
			if len(placeholders) > 0 {
				tokens, uploaded, err := uploadSlidesPlaceholders(runtime, presentationID, placeholders, param)
				if err != nil {
					return appendSlidesProgressHint(err, fmt.Sprintf("presentation %s was created; %d image(s) uploaded before failure", presentationID, uploaded))
				}
				for i := range slides {
					slides[i] = replaceImagePlaceholders(slides[i], tokens)
				}
				result["images_uploaded"] = uploaded
			}

			slideURL := fmt.Sprintf(
				"/open-apis/slides_ai/v1/xml_presentations/%s/slide",
				validate.EncodePathSegment(presentationID),
			)

			var slideIDs []string
			var slideIssues []map[string]interface{}
			for i, slideXML := range slides {
				slideData, err := runtime.CallAPITyped(
					"POST",
					slideURL,
					map[string]interface{}{"revision_id": -1},
					map[string]interface{}{
						"slide": map[string]interface{}{"content": slideXML},
					},
				)
				if err != nil {
					return appendSlidesProgressHint(err, fmt.Sprintf("adding slide %d/%d failed; presentation %s was created, %d slide(s) added before failure", i+1, len(slides), presentationID, i))
				}
				sid := common.GetString(slideData, "slide_id")
				if sid != "" {
					slideIDs = append(slideIDs, sid)
				}
				if issues, ok := slideData["issues"]; ok {
					slideIssues = append(slideIssues, map[string]interface{}{
						"slide_index": i + 1,
						"slide_id":    sid,
						"issues":      issues,
					})
				}
			}

			result["slide_ids"] = slideIDs
			result["slides_added"] = len(slideIDs)
			if len(slideIssues) > 0 {
				result["slide_issues"] = slideIssues
			}
		}

		// Prefer the URL returned by presentation.create. Fall back to a local
		// brand-standard URL only when the API omits it.
		presentationURL := common.GetString(data, "url")
		if presentationURL == "" {
			presentationURL = common.BuildResourceURL(runtime.Config.Brand, "slides", presentationID)
		}
		if presentationURL != "" {
			result["url"] = presentationURL
			if len(slides) == 0 {
				result["message"] = fmt.Sprintf("成功创建空白幻灯片，url：%s，请给用户推送开工通知。", presentationURL)
			}
		}

		if grant := common.AutoGrantCurrentUserDrivePermission(runtime, presentationID, "slides"); grant != nil {
			result["permission_grant"] = grant
		}

		runtime.Out(result, nil)
		return nil
	},
}

// createSlideContents resolves the page XML for +create from whichever input
// form the caller used, and returns it alongside the flag name to blame in
// errors. An empty result means "create an empty presentation".
//
// Two forms, because assembling the JSON array by hand is what callers keep
// getting wrong: --slides takes the finished array (now also from @file/stdin),
// --slide is repeated once per page and the CLI builds the array. The forms are
// mutually exclusive — merging them would make page order depend on flag
// ordering rules nobody wants to reason about.
func createSlideContents(runtime *common.RuntimeContext) ([]string, string, error) {
	// Whether --slides was typed, not whether its value is non-empty: an empty
	// value is the shape a failed command substitution takes
	// (--slides "$(...)"), and silently reading it as "no pages given" is how a
	// caller ends up with a blank deck reported as success. Empty is not valid
	// JSON, so routing it into the parse below rejects it by itself.
	slidesJSON := runtime.Str("slides")
	slidesGiven := runtime.Cmd.Flags().Changed("slides")
	slideArgs := runtime.StrArray("slide")

	if slidesGiven && len(slideArgs) > 0 {
		return nil, "--slide", errs.NewValidationError(errs.SubtypeInvalidArgument,
			"--slide and --slides cannot be combined; pass the whole array with --slides, or one page per --slide").
			WithParam("--slide")
	}

	var (
		slides []string
		param  string
	)
	switch {
	case slidesGiven:
		param = "--slides"
		// A null literal is valid JSON for any slice, so it parses without error
		// and leaves slides nil. Reading that as "no pages given" is the same
		// blank-deck-reported-as-success an empty value would cause, so it is
		// rejected the same way. `[]` stays valid: that array is explicitly empty.
		if err := json.Unmarshal([]byte(slidesJSON), &slides); err != nil || slides == nil {
			return nil, param, errs.NewValidationError(errs.SubtypeInvalidArgument, "--slides invalid JSON, must be an array of XML strings").
				WithParam("--slides").
				WithHint("to pass pages as XML files instead of building the array, repeat --slide @page.xml once per page")
		}
	case len(slideArgs) > 0:
		param = "--slide"
		var err error
		if slides, err = readSlideArgs(runtime, slideArgs); err != nil {
			return nil, param, err
		}
	default:
		return nil, "--slides", nil
	}

	if len(slides) > maxSlidesPerCreate {
		return nil, param, errs.NewValidationError(errs.SubtypeInvalidArgument,
			"%s exceeds maximum of %d slides per create; create the presentation first, then add the remaining pages one at a time with slides +add-slide",
			param, maxSlidesPerCreate).WithParam(param)
	}

	// Structural checks run on the assembled array, so both input forms fail the
	// same way. The backend rejects a non-<slide> payload with an opaque 3350001
	// after the presentation already exists; catching it here keeps the deck from
	// being created at all.
	for i, slideXML := range slides {
		slides[i] = strings.TrimSpace(slideXML)
		if slides[i] == "" {
			return nil, param, errs.NewValidationError(errs.SubtypeInvalidArgument, "%s: page %d is empty", param, i+1).WithParam(param)
		}
		if err := validateCompleteSlideXML(slides[i]); err != nil {
			return nil, param, errs.NewValidationError(errs.SubtypeInvalidArgument,
				"%s: page %d is not a single complete <slide> document: %v", param, i+1, err).WithParam(param).WithCause(err)
		}
	}
	return slides, param, nil
}

// readSlideArgs turns each --slide value into page XML, reading @path values
// from disk. The framework only resolves Flag.Input for single-valued string
// flags, so a repeatable flag has to do it here — deliberately through the same
// reader, which enforces the "relative path under the current directory" rule.
func readSlideArgs(runtime *common.RuntimeContext, args []string) ([]string, error) {
	slides := make([]string, 0, len(args))
	for i, raw := range args {
		raw = strings.TrimSpace(raw)
		switch {
		case raw == "-":
			// A process has one stdin, so "-" cannot mean "this occurrence" on a
			// repeatable flag. --slides reads stdin as the whole array.
			return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "--slide does not support stdin (-)").
				WithParam("--slide").
				WithHint("pass each page as --slide @page.xml, or pipe the whole JSON array into --slides -")
		case strings.HasPrefix(raw, "@@"):
			// Same escape the framework applies: @@ is a literal leading @.
			slides = append(slides, raw[1:])
		case strings.HasPrefix(raw, "@"):
			path := strings.TrimSpace(raw[1:])
			if path == "" {
				return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "--slide: file path cannot be empty after @ (page %d)", i+1).WithParam("--slide")
			}
			data, err := cmdutil.ReadInputFile(runtime.FileIO(), path)
			if err != nil {
				return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "--slide: %v", err).WithParam("--slide").WithCause(err)
			}
			// Same normalization the framework applies to Input flags: a
			// repeatable flag resolves @path itself, so it has to strip the BOM
			// itself too, otherwise --slide @page.xml rejects a file that
			// --slides @deck.json accepts.
			slides = append(slides, common.StripUTF8BOM(string(data)))
		default:
			slides = append(slides, raw)
		}
	}
	return slides, nil
}

// effectiveTitle returns the title to use, falling back to "Untitled".
func effectiveTitle(title string) string {
	if title == "" {
		return "Untitled"
	}
	return title
}

// buildPresentationXML builds the minimal XML for a new empty presentation.
func buildPresentationXML(title string) string {
	escapedTitle := xmlEscape(title)
	if escapedTitle == "" {
		escapedTitle = "Untitled"
	}
	return fmt.Sprintf(
		`<presentation xmlns="https://www.larkoffice.com/sml/2.0" width="%d" height="%d"><title>%s</title></presentation>`,
		defaultPresentationWidth, defaultPresentationHeight, escapedTitle,
	)
}

// uploadSlidesPlaceholders uploads each unique placeholder path against the
// presentation and returns the path→file_token map. The second return value is
// the number of files successfully uploaded before any error, so callers can
// surface progress in the failure message. param names the flag the XML came
// from, so an error points at the flag the caller actually typed.
func uploadSlidesPlaceholders(runtime *common.RuntimeContext, presentationID string, paths []string, param string) (map[string]string, int, error) {
	tokens := make(map[string]string, len(paths))
	for i, path := range paths {
		stat, err := runtime.FileIO().Stat(path)
		if err != nil {
			return tokens, i, slidesInputStatError(err, param, fmt.Sprintf("@%s", path))
		}
		if !stat.Mode().IsRegular() {
			return tokens, i, errs.NewValidationError(errs.SubtypeInvalidArgument, "@%s: must be a regular file", path).WithParam(param)
		}
		fileName := filepath.Base(path)
		fmt.Fprintf(runtime.IO().ErrOut, "Uploading image %d/%d: %s (%s)\n",
			i+1, len(paths), fileName, common.FormatSize(stat.Size()))

		token, err := uploadSlidesMedia(runtime, path, fileName, stat.Size(), presentationID)
		if err != nil {
			return tokens, i, fmt.Errorf("@%s: %w", path, err) //nolint:forbidigo // intermediate; preserves typed cause via %w, reclassified by appendSlidesProgressHint at the call site
		}
		tokens[path] = token
	}
	return tokens, len(paths), nil
}

// xmlEscape escapes special XML characters in text content.
func xmlEscape(s string) string {
	s = strings.ReplaceAll(s, "&", "&amp;")
	s = strings.ReplaceAll(s, "<", "&lt;")
	s = strings.ReplaceAll(s, ">", "&gt;")
	s = strings.ReplaceAll(s, "\"", "&quot;")
	s = strings.ReplaceAll(s, "'", "&apos;")
	return s
}