Fix DictsView.dicts instance attr shadowing Table.dicts() method

[gaoflow]

Problem

fromdicts() returns a DictsView (in petl/io/json.py) that stores its
input data as self.dicts. That name clashes with the dicts() method
inherited from Table, which is bound at the class level via
Table.dicts = dicts in petl/util/base.py.

The instance attribute wins in Python's attribute-lookup order, so any
.dicts() call on the result raises:

TypeError: 'DictsView' object is not callable

Minimal reproducer (reported in #643):

import petl as etl
dummy = etl.dummytable()
table = etl.fromdicts(dummy.dicts())
table.dicts()   # TypeError: 'DictsView' object is not callable

Fix

Rename the internal attribute self.dicts → self._dicts in both
DictsView and DictsGeneratorView, consistent with the existing
self._header convention. Five call-sites updated (two in
__init__, one in DictsView.__iter__, two in
DictsGeneratorView.__iter__ / _determine_header).

A regression test is added to petl/test/io/test_json.py.

Fixes #643.

---

This pull request was prepared with the assistance of AI, under my direction and review.

[qodo-free-for-open-source-projects]

PR Summary by Qodo

Fix fromdicts() DictsView attribute shadowing inherited Table.dicts()
🐞 Bug fix 🧪 Tests 🕐 10-20 Minutes

Description

• Rename DictsView internal storage from dicts to _dicts to avoid method shadowing.
• Ensure .dicts() on fromdicts() results returns row dictionaries instead of raising TypeError.
• Add regression test covering the .dicts() call path for DictsView.

Diagram

graph TD
  A["User code"] --> B["etl.fromdicts()"] --> C["DictsView / DictsGeneratorView"] --> D["iterdicts()"]
  C --> E["Table.dicts() method"]
  subgraph Legend
    direction LR
    _fn["Function"] ~~~ _cls["Class/View"]
  end

Loading

High-Level Assessment

The following are alternative approaches to this PR:

1. Rename Table.dicts() to a less collision-prone name

• ➕ Eliminates a broad class of potential name collisions with instance attributes named dicts
• ➖ Breaking API change across the library and downstream users
• ➖ Requires deprecation cycle and widespread updates

2. Convert DictsView storage to a property with a different backing name

• ➕ Keeps readable attribute access while avoiding collisions (e.g., dicts property backed by _dicts)
• ➖ Still risks confusion with the method name dicts()
• ➖ Adds unnecessary indirection for an internal-only field

Recommendation: Proceed with the PR’s approach: renaming the internal attribute to _dicts is the smallest, safest change, consistent with existing private naming (e.g., _header) and avoids any public API churn. Alternatives either introduce breaking changes (renaming Table.dicts()) or add complexity without improving clarity (property indirection).

Files changed (2) +21 / -5

Bug fix (1) +5 / -5

json.pyAvoid shadowing Table.dicts() by renaming DictsView storage to '_dicts' +5/-5

Avoid shadowing Table.dicts() by renaming DictsView storage to '_dicts'

• Changes DictsView and DictsGeneratorView to store input data in 'self._dicts' instead of 'self.dicts'. Updates iteration and header-determination code paths to reference '_dicts', preventing '.dicts()' from being shadowed by an instance attribute.

petl/io/json.py

Tests (1) +16 / -0

test_json.pyAdd regression test for calling '.dicts()' on fromdicts() results +16/-0

Add regression test for calling '.dicts()' on fromdicts() results

• Adds a test ensuring 'fromdicts(...).dicts()' returns the expected row dictionaries and does not raise 'TypeError'. Covers the specific regression caused by attribute/method name collision.

petl/test/io/test_json.py

[qodo-free-for-open-source-projects]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0) 🎨 UX issues (0) 🔗 Cross-repo conflicts (0) 📜 Skill insights (0)

1. test_fromdicts_dicts_method lacks dummytable ✓ Resolved 📎 Requirement gap ≡ Correctness

Description

The new regression test calls fromdicts() with a literal list of dicts, but it does not reproduce
the reported sequence dummy = etl.dummytable(); table = etl.fromdicts(dummy.dicts()). This may
miss regressions specific to consuming dummy.dicts() (a generator/view) and does not meet the
checklist’s explicit success criteria.

Code

petl/test/io/test_json.py[R344-357]

+def test_fromdicts_dicts_method():
+    # fromdicts() returns a DictsView whose internal data attribute was named
+    # 'dicts', shadowing the inherited Table.dicts() method.  Calling .dicts()
+    # on the result should return row dicts, not raise TypeError.
+    data = [{'foo': 'a', 'bar': 1},
+            {'foo': 'b', 'bar': 2},
+            {'foo': 'c', 'bar': 2}]
+    actual = fromdicts(data, header=['foo', 'bar'])
+    result = list(actual.dicts())
+    assert result == [
+        {'foo': 'a', 'bar': 1},
+        {'foo': 'b', 'bar': 2},
+        {'foo': 'c', 'bar': 2},
+    ]

Evidence

PR Compliance ID 2 requires a test that uses etl.dummytable() and etl.fromdicts(dummy.dicts()).
The added test instead constructs data = [...] and calls actual = fromdicts(data, ...), which
does not match the required reproducer steps.

Add regression coverage for DictsView/method name collision on dicts
petl/test/io/test_json.py[344-357]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The regression test does not follow the required reproducer sequence (`dummy = etl.dummytable(); table = etl.fromdicts(dummy.dicts()); table.dicts()`), so it may not prevent reintroduction of the specific collision scenario described in the checklist.
## Issue Context
Checklist PR Compliance ID 2 explicitly requires using `etl.dummytable()` and passing `dummy.dicts()` into `etl.fromdicts(...)`, then asserting `table.dicts()` is callable and yields dict rows.
## Fix Focus Areas
- petl/test/io/test_json.py[344-357]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

[gaoflow]

Thanks, that is a good catch. I pushed 4ba9c84 to make the regression test follow the reported path exactly:

dummy = dummytable(numrows=3, seed=42)
actual = fromdicts(dummy.dicts())
list(actual.dicts())

Local verification:

uv run --with pytest --with pytest-cov --with-editable . python -m pytest petl/test/io/test_json.py::test_fromdicts_dicts_method -q
uv run --with pytest --with pytest-cov --with-editable . python -m pytest petl/test/io/test_json.py -q
uv run --with ruff ruff check petl/test/io/test_json.py petl/io/json.py
git diff --check

I also checked the current red 3.8/3.9 CI jobs. They fail before pytest runs, while building the package via python setup.py sdist bdist_wheel, with AttributeError: ignore_egg_info_in_manifest from the setuptools/setuptools_scm integration in the runner environment. That does not look related to this test/code path.