[saudsami]

Reported against setRemoteVideoStream(_:type:): the streamType parameter names AgoraVideoStreamType but doesn't link to it, where the Android page links the equivalent type.

It isn't one page

The iOS template contains zero DocC symbol links. No API in the Apple references links to any other.

xml2json.py resolves an <xref> to its display text and discards the link:

xref.text = href_text      # line 733 — display text kept, link lost

Nothing downstream recovers it: iris_doc's fmt_oc.yaml declares link1: ""/link2: "", but those fields are never read in its production code — they appear only in test fixtures. The Oxygen-built platforms are unaffected because they render <xref> as an <a href> straight from the DITA, which is why Android looks right.

The change

Wrap resolved targets in DocC symbol-link syntax on the Markdown platforms. Because the substitution happens on the element tree before extraction, this reaches parameters, discussions and return values without touching any of those code paths.

A target containing a space isn't a symbol name (technical support), so it stays plain rather than rendering as a meaningless code span.

Scope

Measured against the 3,860 Swift and Objective-C names in the shipped iOS archive:

cross-references reaching the iOS docs	3,511
target exists in the iOS SDK — will link	3,009 (86%)
not a symbol — left plain	230
keyref has no keydef (pre-existing source gap)	272

Verification

iOS template	2,009 symbol links, from 0
unity, flutter, rn, electron	byte-identical to master

The reported case:

streamType: 'The video stream type, see ``AgoraVideoStreamType``.'

Known limitation

A `` wrapper around a symbol DocC can't resolve renders as a code span, not a link. With 86% resolvable, expect a few hundred of those — an improvement on plain prose, but not a link. The 272 keyrefs with no keydef remain broken at source and produce empty text today; worth a separate look.

Worth confirming with a real build before this ships, as the template alone can't show whether DocC resolves each target.

🤖 Generated with Claude Code

[Foesa-yyx]

Thanks for addressing the missing DocC link. One gap remains: the new symbol_link helper is only used in the keyref and href-with-fragment branches (xml2json/xml2json.py:751 and :783). The no-fragment branch still assigns xref.text = title_text.strip() at line 810, so a reference such as <xref href="class_livetranscoding.dita" /> continues to generate plain See AgoraLiveTranscoding. on iOS/macOS. Please pass this value through symbol_link(title_text.strip(), platform_tag) and add a regression case for file-level hrefs. The reported AgoraVideoStreamType keyref case is fixed.

[saudsami]

Good catch — confirmed and fixed in 651e0910.

You're right that there are three branches resolving an <xref> to display text (keyref, href-with-fragment, href naming a file) and only the first two went through symbol_link. Line 810 now does too:

xref.text = symbol_link(title_text.strip(), platform_tag)

Impact: three file-level references reach the iOS docs; one more link is emitted (2,009 → 2,010), including your AgoraLiveTranscoding example, which now resolves on updateRtmpTranscodingEx.

Regression case

Added xml2json/test_xml2json.py — the first tests in this directory:

• SymbolLinkTest — symbol_link behaviour: Markdown platforms, other platforms unchanged, the space guard for prose targets like technical support, empty targets, Objective-C selectors.
• XrefResolutionTest — asserts every branch assigning xref.text routes through symbol_link, plus a specific case for the file-level branch.

I pointed the structural test at the failure mode you found rather than just the one line, so a fourth branch added without the helper fails too. The http branch is exempt — it appends a URL rather than naming a symbol.

Verified the test actually catches it: 2 failures with the fix reverted, clean with it in place. unity and flutter remain byte-identical to master.

python3 -m unittest discover -s xml2json -p 'test_*.py'
Ran 7 tests — OK

One thing worth flagging separately, found while scoping this: 272 keyrefs in the iOS maps have no keydef, so they resolve to empty string today. That's why setVideoEncoderConfiguration reads "Encoder: Use to select H.265" with the API name missing. Source-side rather than converter-side, and not addressed here.

🤖 Generated with Claude Code

[Foesa-yyx]

@saudsami I’ve merged the PR. Thanks for the fix and for addressing the review feedback!