Response shapes

The four shapes

data is not one thing. Check what type it is before deciding how to read it:

Shape	Type of data	How to read it
Row array	array, one item per row	data[0].field; more rows are data[0] → data[n]
Table list	array of arrays, one item per table	data[0] is the first table, data[0][0] is its first row
Pre-rendered text	string (markdown)	use data for display; use the extra structured / tables for programmatic access
Multi-segment structure	object containing tables	read segments from data.tables; each has title / slug / rows

A one-line rule: look at whether the first character is [, { or ".

Which of the 33 endpoints uses which shape

Shape	Endpoints
Row array	quote · kline · minute · hot · technical · chip · profile · search · ipo · dividend · block-trade · margin-trade · orderbook · snapshot
Table list	finance
Pre-rendered text	market-overview · changedist · events · lhb · sector · sector-valuation · macro · bond · etf · dehydrated · screen
Multi-segment structure	shareholder · calendar
Object (neither array nor text)	fund · industry-chain
Special cases	health (no data) · usage (plain text, not JSON) · challenge (returns the challenge itself)

The ones that trip people up

orderbook and snapshot are row arrays too, just always exactly one row. The design keeps everything a row array so clients do not need a special case for single-object endpoints — read data[0].field and you are right.

⚠️ The table above is the shape for a single code. Sending several codes changes the shape — quote / technical / profile add a code column, while chip becomes a wide table with prefixed column names (the row count does not double). See Batch fetching for details.

finance is a table list, not a multi-segment structure. data[0] / data[1] / data[2] are the income statement / balance sheet / cash-flow statement — the order is fixed. To address them by name, see the next section.

shareholder and calendar are multi-segment structures. data.tables is an ordered list of tables. ⚠️ Do not pick shareholder segments by index — A-shares and Hong Kong both have 3 segments, but they mean completely different things (A-shares: top-ten holders / top-ten float holders / holder-count stats; Hong Kong: shareholders / distribution / institutional holdings). Address them by slug.

fund is a genuine object (not text): read data.main_net, data.main_rank and so on directly.

How to read text endpoints programmatically

For text endpoints, data is human-readable markdown, but two extra fields appear alongside:

Field	What it is
structured	The first table, as [{column: value}]
tables	Every table, as [[{column: value}], ...]

data itself is unchanged — text clients keep using data, programmatic clients use structured / tables. For multi-segment output (such as calendar or macro), only tables gets you everything.

structured gives you just the first table, so use it when one table is all you need (e.g. the screen result); when there are several segments, always use tables.

Common misuses

• Treating a text endpoint as an array — for market-overview, sector and friends data is a string, so data[0] is the first character, not the first row.
• Treating a row array as an object — quote’s data is an array; data.last finds nothing, you need data[0].last.
• Addressing finance by name — it is an order-fixed table list with no slugs; addressing by name gets you the wrong table.
• Picking shareholder segments by index — A-shares and Hong Kong have the same segment count but different meanings; always use slug.
• Forgetting structured holds only the first table — reading only it silently drops every later segment.