Fill 17,252 empty exchange + market cells in equities.csv

[dokson]

Summary

Fills 17,252 of the 17,338 empty exchange + market cells in equities.csv — and the consequence-of-#143 invariant test (test_exchange_market_one_to_one) still passes on the result.

Pass	Method	Cells filled
1	Deterministic, from ticker suffix using FD's own data	17,240
2	External resolution (yfinance) for US-style symbols	12
Total exchange filled		17,252
Total market filled (1:1 with exchange)		17,252

No row with a pre-existing populated exchange was modified. Only empty cells filled.

1. Deterministic pass (17,240 rows, no external API)

For every row with exchange empty AND a recognisable ticker suffix (.NX, .F, .DU, .MI, …), the canonical exchange code is derived from rows that already have that same suffix populated in FD. Every suffix maps to exactly one exchange code at ≥99% in the existing data.

Suffix	Maps to (FD canonical)	Rows filled
.NX	ENX (Euronext)	11,364
.F	FRA (Frankfurt Stock Exchange)	3,068
.DU	DUS (Dusseldorf Stock Exchange)	624
.BE	BER (Berlin Stock Exchange)	370
.MI	MIL (Borsa Italiana)	349
.MU	MUN (Munich Stock Exchange)	247
.SG	STU (Stuttgart Stock Exchange)	218
.OL	OSL (Oslo Bors)	115
.TI	TLO (EuroTLX)	115
.ST	STO (NASDAQ OMX Stockholm)	89
.BO	BSE (BSE India)	75
.PA	PAR (Euronext Paris)	70
.VI	VIE (Vienna Stock Exchange)	68
.MX	MEX (Mexico Stock Exchange)	63
Other long-tail suffixes	various	~405

The market column is filled from FD's own canonical exchange → market mapping (also derived from existing populated rows).

2. External resolution pass (12 rows, US-style symbols)

After Pass 1, 98 rows still had empty exchange. These are US-style symbols (no ticker suffix) with every other column also NaN — failed scrapes from an older import. yfinance resolved 12:

3. The 86 remaining rows — out of scope here

After both passes, 86 rows still have empty exchange. All are US-style symbols representing delisted tickers (old SPAC units like CCX'U, warrants like BFT-WT, post-bankruptcy Q-suffixed tickers like SVTLQ, etc.).

Cross-checked all 86 against Morningstar's published Re-Used Ticker Symbols list:

• 9 tickers are listed as reused (the symbol has been reassigned to a different company by an exchange): CATS, CEL, CTL, ENT, ERI, HDS, MGEN, RMG, WMGI. These are ambiguous — the FD row represents the historical company but the live symbol now belongs to a different one. Safer left empty than written with a current exchange that would mislead consumers about which entity the row describes.
• 77 tickers are NOT in the reused list (truly delisted, never reassigned). These are perfect candidates for the delisted.csv file proposed in [IMPROVE] Split equities/etfs/funds CSVs by exchange for maintainability (no data change) #144 (per-exchange shards + a single delisted file as a clean partition of "currently listed").

Validation

Passes the test_exchange_market_one_to_one invariant introduced in #143:

Each exchange code must map to exactly one market label.

Every newly-filled exchange/market pair was sourced from FD's own canonical mapping, so the invariant cannot be broken by this PR.

Diff shape

1 file changed, 17,253 insertions / 17,253 deletions on database/equities.csv. Two columns touched (exchange, market); no rows added or removed.

Test plan

• CI categorization + markdown linters pass
• pytest tests/ — all tests pass; in particular test_exchange_market_one_to_one passes
• Spot-check: a .NX ticker (e.g. XYWG.NX) should now have exchange = ENX, market = "Euronext"
• Spot-check: an .F ticker should now have exchange = FRA, market = "Frankfurt Stock Exchange"
• Confirm no row that already had exchange populated has been modified

Related: #143 (introduced the invariant test), #144 (proposes the delisted.csv home for the 77 remaining rows).