You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 2eb2e62
Browse filesBrowse the repository at this point in the historyBrowse files
Stefan Jansen
committed
docs: identify released migration inputs and consumer boundaries
Correct stale unreleased wording for Data 0.3.0, link release examples, and distinguish historical integration checks from released Data availability. Keep Diagnostic analytics and Backtest runtime/account responsibilities explicit.
Closes#137
Clean published-wheel installation checks passed all six native examples; the optional conversion
11
+
passed 13 checks using the real Zipline readers and writers. The pinned legacy environment and
12
+
historical cross-library checks are described in the relevant sections below. They do not qualify
13
+
every version of a consuming library.
14
+
15
+
| Input needed by a consumer | Released Data API | Boundary |
16
+
|---|---|---|
17
+
| A pandas price panel or native long price table for Diagnostic |`to_price_panel`, `from_price_panel`| Diagnostic computes returns, factor statistics and analytical fill views. Data retains missing quotes and stable IDs. |
18
+
| Captured action facts and dated adjusted research prices |`prepare_action_inputs`, `prepare_adjusted_history`, `InstrumentRegistry`| Backtest owns runtime history, orders, entitlement, payments and positions. Adjusted research prices do not establish broker accounting. |
19
+
| Individual futures definitions, eligible fields and paired roll marks |`prepare_futures_inputs`| Backtest owns roll execution, settlement and margin. Data does not infer missing execution marks. |
20
+
| Portable captured tables and declarations |`export_snapshot`, `read_snapshot`| Each consumer supplies its own effective configuration and runtime policy. |
21
+
| Legacy acquisition imports or a saved Zipline bundle |`ml4t.data.compat.empyrical`, `ml4t.data.compat.zipline`| These adapters preserve the bounded data workflow, not the legacy analytics or execution engine. |
22
+
23
+
The examples qualify Data's producer contracts. For a combined environment, also use the
24
+
consumer's published migration guide and qualification evidence. The historical source checks
25
+
below remain limited to their named revisions; they are not current installation requirements.
26
+
3
27
## Move a long price table to a pandas analytical panel
4
28
5
29
Keep source observations in Data storage. Use `to_price_panel` when an analytical consumer needs
@@ -56,7 +80,7 @@ the declared label clock; instant bounds use UTC. Read ranges are start-inclusiv
56
80
Calendar omission, explicit `None` and a named `24/7` schedule remain distinct. Panel conversion
57
81
does not invoke a calendar resolver or create an expected clock.
58
82
59
-
These adapters are unreleased source functionality. Isolated wheel and source-distribution checks
83
+
Before the 0.3.0 release, isolated wheel and source-distribution integration checks
60
84
qualified Data `0.2.2.dev14+gbe73c5d86`, Specs `0.2.0`, Diagnostic `0.1.9` and Engineer
(`0.1.13.dev22+g12a2c483f`) on Python `3.12.11`. Actual Backtest `ContractSpec` and `Position`
386
410
consumed the produced raw and converted fields, retaining IDs, paired marks and units; both
@@ -389,7 +413,7 @@ roll fills, settlement, margin or a complete released-package migration.
389
413
Data does not replace Backtest's runtime history, account, roll execution or auto-close policy.
390
414
The recipe needs no credentials, network, Reloaded package or Backtest installation.
391
415
392
-
This source candidate requires Polars `>=1.36.1` and pandas-market-calendars `>=5.4.0`.
416
+
Data 0.3.0 requires Polars `>=1.36.1` and pandas-market-calendars `>=5.4.0`.
393
417
Isolated runs using the former advertised minima failed during ordinary Data import or calendar
394
418
construction. The saved-file recipe passes with Polars `1.36.1`, pandas-market-calendars `5.4.0`
395
419
and pandas `2.3.3`; those checks do not certify every possible dependency-version combination.
@@ -457,12 +481,13 @@ from ml4t.data.compat.empyrical import default_returns_func
457
481
register_return_func(default_returns_func)
458
482
```
459
483
460
-
The offline [acquisition migration example](https://github.com/ml4t/data/blob/3dd71cb394c390316ff0bb36ad1d7d0f3dd6aefc/examples/acquisition_migration.py)
484
+
The offline [acquisition migration example](https://github.com/ml4t/data/blob/db9cda776b5b346f536b0e8ceb855b9fd60b6ec8/examples/acquisition_migration.py)
461
485
replays retained public response captures through actual parsers, providers and CSV files. It runs
462
-
without credentials or network and does not require any Reloaded package. This is an unreleased
463
-
source workflow, not a claim that a released successor stack already replaces every legacy option.
486
+
without credentials or network and does not require any Reloaded package. This workflow is
487
+
included in Data 0.3.0; it does not establish replacement of every legacy option.
464
488
465
-
Isolated wheel and source-distribution checks qualified Data `0.2.2.dev27+g5ff44e4ac`,
489
+
Pre-release wheel and source-distribution integration checks qualified Data
490
+
`0.2.2.dev27+g5ff44e4ac`,
466
491
Specs `0.2.0`, Diagnostic `0.1.9`, Pyfolio Reloaded `0.10.0` at immutable commit
467
492
`365e649f591c39cef5d0170958280e49b8eeb953`, and Alphalens Reloaded `0.5.0` at
468
493
`1bc181de4aafbe33437ed0e362ff96aa7ce20acd`. The checks exercised actual Diagnostic
@@ -476,7 +501,7 @@ the Yahoo extra now requires the qualified `>=1.5.2` range.
476
501
## Reconstruct inputs after source pruning
477
502
478
503
Retain the input tables and declarations instead of referring to mutable source paths. Run the
0 commit comments