sync dev -> main - #2935
Draft
ValueRaider wants to merge 43 commits into
Draft
Conversation
…ner fields The Yahoo Finance screener API supports both `dividendyield` (trailing yield percentage) and `dividendpershare.lasttwelvemonths` (trailing DPS) as query/filter/sort fields. These were missing from the profitability category in EQUITY_SCREENER_FIELDS, causing EquityQuery to reject queries referencing them. Forward equivalents (forward_dividend_yield, forward_dividend_per_share) were already present.
…r-field fix: add dividendyield and dividendpershare.lasttwelvemonths to screener fields
…tments, NetLoan, DeferredAssets
pd.Timedelta() was called with bare interval strings (e.g. "1h", "30m"), which numpy>=2.5 flags with 'The generic unit for NumPy timedelta is deprecated'. Parse minute/hour intervals with explicit units in _interval_to_timedelta and route _dts_in_same_interval through it so both warning sites are covered. Fixes #2882 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Lookup class stores the search term as self.query, but two error-handling paths in _fetch_lookup referenced self.ticker, which does not exist on the class. When Yahoo returns a finance error payload (or faulty JSON), the intended YFDataException / error log is never produced; instead the code raises a confusing 'Lookup object has no attribute ticker' AttributeError that masks the real cause. Replace both self.ticker references with self.query so the error is reported as intended. Add an offline unit test that feeds a mocked finance-error response and asserts YFDataException is raised with the query in the message.
Add missing balance sheet keys: FixedMaturityInvestments, EquityInves…
Fix generic-unit timedelta DeprecationWarning for sub-day intervals (#2882)
Series.to_numpy() can return a read-only view under Pandas copy-on-write, so the in-place reduction of huge 'Adj Close' values (added in #2860) crashed with 'ValueError: assignment destination is read-only' e.g. for ENV.CR. Copy the array before modifying; it is written back to df2 afterwards anyway. Add offline regression test. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
yfinance works around a Yahoo API bug by fetching 15m data and resampling it into 30m bars. That internal 15m interval leaked into YFPricesMissingError messages: a failed 30m request reported '15m data not available', confusing users who never asked for 15m (reported 2022, still reproducible on 1.4.1). - Build the error context from the interval the user requested. - When Yahoo's quoted error names the internally fetched interval, append a note explaining the substitution. - Document the 30m-to-15m mapping in history() and download() docstrings, which the docs website is generated from. - Add live regression test that fails before this change. Closes #1029 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
fix: read-only 'Adj Close' array crash in dividend-adjust repair
…-ticker Fix AttributeError in Lookup error handling (self.ticker -> self.query)
Use the phrasing requested by ValueRaider in PR #2900: '(30m resampled from 15m)'. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…-message Fix error messages showing internal 15m interval for 30m requests
pandas 3 attaches zoneinfo timezones to DatetimeIndex, so comparing
against a pytz object fails on any pandas >= 3:
AssertionError: zoneinfo.ZoneInfo(key='America/New_York')
!= <DstTzInfo 'America/New_York' LMT-1 day, 19:04:00 STD>
Compare the timezone key string instead, which is identical for the
zoneinfo (pandas 3) and pytz (pandas < 3) representations, matching
the pandas floor of >= 1.3 in requirements.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Every YFPricesMissingError message speculates 'possibly delisted', even when Yahoo's response carries an explicit reason that has nothing to do with delisting, e.g. an intraday range outside the last 60 days. That speculation has misled users for years (see #1713, #1797, #2044, #2052). And when a ticker really is dead, Yahoo's quoted description already says 'No data found, symbol may be delisted', making the prefix redundant there too. Add possibly_delisted=True parameter to YFTickerMissingError and YFPricesMissingError (default keeps every other callsite unchanged) and pass False in the one history() branch where Yahoo supplied an error description, letting Yahoo's stated reason stand. Adds an offline test of both message formats and a live test that a range error no longer claims delisting. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…ndas3-tz Fix pandas 3 timezone assertion in test_download_multi_small_interval
Per review on PR #2903: when Yahoo explains why data is missing, drop the remaining boilerplate too, not just the 'possibly delisted' prefix. The message is now just the ticker and Yahoo's reason: $SPY: 1m data not available for startTime=... Only 8 days worth of 1m granularity data are allowed to be fetched per request. $SPSPS: No data found, symbol may be delisted Replace the possibly_delisted flag on YFPricesMissingError with a yahoo_reason argument that becomes the whole message. The generic 'no price data found' text and the request-context wrapper are dropped for this path; the request context is still kept on the .debug_info attribute. The #2900 resample note still appends when the fetched interval differs, so a 30m failure does not regress to reading as 15m. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…aim-on-yahoo-error Don't claim 'possibly delisted' when Yahoo explains the missing data
Fix: handle null result from Yahoo API in _fetch_info
docs: fix grammar and formatting in advanced documentation
- calibration resilient against sudden-changes in df_new - better calibration of 'Adj Close' in df_new - repair invalid OHLC e.g. Close<Low - Use Volume to decide if a sudden-change is unit-switch or missing-split - Fix intraday div-repair by resampling to 1d then backporting repairs - Handle more bad-div scenarios - correct_columns_individually auto-set according to data - recalc invalid Low/High - unit-switch repair updates currency in history_metadata - Improve handle nan-prices with dividend
Re-enable test_no_expensive_calls_introduced without requests_cache
Price repair fixes & improvements
ruff 0.16.0 expanded the default rule set, so the unpinned, config-less lint job started failing on pre-existing code across the repo. Add ruff.toml pinning select to E4,E7,E9,F (ignore E702, exclude the generated pricing_pb2.py) and simplify the workflow args. No code changes.
Pin ruff ruleset for deterministic CI
Move build config and metadata from setup.py/setup.cfg to pyproject.toml so [project.dependencies] is the single source of truth, and switch the publish workflow to python -m build. Also declare lxml (needed by pandas.read_html), add Python 3.11-3.13 classifiers, and drop the unusable nospam extra.
On pandas 2.x with numpy>=2.5, constructing a Timedelta from a bare string or from a keyword like days=/minutes= emits the "'generic' unit for NumPy timedelta is deprecated" DeprecationWarning; only the (value, unit=...) form is silent. An earlier fix switched the sub-day branches of _interval_to_timedelta to minutes=/hours=, but that keyword form still warns, so the warning persisted. Use the (value, unit=...) form there and in Quote.shares.
…rvals _interval_to_timedelta returns a relativedelta for day intervals, which cannot be compared against the Timedelta (dt2 - dt1), so _dts_in_same_interval raised TypeError for a multi-day interval such as "5d". Build a Timedelta directly for that case, using the (value, unit="D") form so it does not reintroduce the generic-unit deprecation warning.
Migrate packaging to pyproject.toml
… main currency _standardise_currency() converts subunit currencies (pence, cents, agora) to main currency internally to help repair dividend-adjustment math, but this conversion was leaking into the final returned data and metadata, even when no actual repair was needed. Now reverts prices/dividends/currency back to original units before returning to the user. Verified against XDEV.L, HSBA.L (incl. auto_adjust), NPN.JO, and AAPL. Confirmed bug reproduces on unmodified dev and is fixed on this branch. Ran tests/test_price_repair.py: 2 pre-existing failures (test_resampling, test_repair_zeroes_daily) reproduce identically on unmodified dev, unrelated to this change. - Refactored currency scaling dict into _CURRENCY_CONVERSIONS module-level constant - Added regression test for issue #2866 Fixes #2866
Fix: repair=True should not permanently convert GBp/ZAc/ILA prices to main currency
Fix: NaN equality comparison
Migrate to pyproject.toml
Fix multi-day interval TypeError and actually silence the numpy>=2.5 Timedelta warning
download()'s ignore_tz docstring explained the parameter's effect on combining timezones, but never stated what timezone the returned index actually carries in either case. That's a real gap: since 1.4.0 (#2825, "Fix localized intraday download() always returning UTC"), ignore_tz=False (the intraday default) tz-converts to the most common exchange timezone among the requested tickers, not UTC as before - correct and changelog'd, but an upgrade a caller pinning old behavior would only discover by diffing output, since the docstring itself never said what to expect either before or after. Verified against reindex_dfs(): tz_localize(None) when ignore_tz=True, tz_convert(tz_mode) (mode of each df's index.tz) when False. No behavior change, docstring only.
docs: state download()'s actual returned index timezone contract
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
12 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
It's time ...