fix(pretty_export): stringify related objects when joining list values by Sanjays2402 · Pull Request #1112 · smithyhq/sqladmin

Sanjays2402 · 2026-07-31T05:34:30Z

PrettyExport._base_export_cell joined a relationship's formatted value with ",".join(formatted_value), but for an InstrumentedList the default formatter returns the related model instances themselves, so exporting a column_list containing a to-many relationship raised TypeError: sequence item 0: expected str instance, Address found. Each item is now coerced with str() before joining, matching how the non-list relationship branch renders its value.

Regression test added in tests/test_pretty_export.py beside the existing _get_export_row_values tests; it fails with the reported TypeError without the fix and passes with it.

This change was prepared with AI assistance; the regression test was run locally and fails without the fix.

_base_export_cell joined a relationship's formatted value with ",".join(formatted_value). The default formatter returns the related model instances themselves for an InstrumentedList, so exporting a column_list containing a to-many relationship raised TypeError: sequence item 0: expected str instance, Address found. Each item is now coerced with str() before joining, matching how the non-list relationship branch renders its value. Closes smithyhq#1111

Copilot

Pull request overview

Fixes PrettyExport._base_export_cell to correctly export to-many relationship values by coercing each formatted item to str before joining, preventing TypeError when the default formatter returns related model instances (e.g., SQLAlchemy InstrumentedList of Address objects).

Changes:

Update relationship list rendering to ",".join(str(item) for item in formatted_value) to avoid joining non-strings.
Add a regression test covering exporting a column_list that includes a to-many relationship.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
`sqladmin/pretty_export.py`	Coerces relationship list items to strings before joining to prevent export-time `TypeError`.
`tests/test_pretty_export.py`	Adds regression test verifying relationship list export no longer errors and returns a joined string.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

mmzeynalli · 2026-08-04T08:40:16Z

+            assert len(values) == 3
+            assert values[0] == 1
+            assert values[1] == "John Doe"
+            assert values[2] == ",".join(str(address) for address in user.addresses)


Add __str__ to Adress

mmzeynalli · 2026-08-04T15:03:14Z

One more thing. This PR considers relationships rendered as QuerySelectField (scalar — uselist=False, a plain many-to-one). It does not fix relationships rendered as QuerySelectMultipleField.

column_import_list = [User.name, User.addresses]
CSV:  name,addresses
      Bob,adg34gfb13

branch result: {"ok": true, "imported": 1, "skipped": 0, "missed_rows": []}

This PR works by keeping errors that fallback_form.validate() already reported. It depends on WTForms actually reporting one. For a collection relationship, validate() returns True with errors == {} even though the value is garbage. There is no error to keep, so the new block at lines 130–135 has nothing to copy and falls through silently.

…relationships

Copilot

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

Suppressed comments (1)

sqladmin/pretty_export.py:33

The iterable-join decision is based on formatted_value, not the underlying relationship value. This can change behavior for to-one relationships if a custom formatter returns an iterable (e.g., dict/list/tuple): it will now be comma-joined instead of being passed through. Also, joining sets without sorting yields nondeterministic output order. Consider joining only when the relationship value is a collection, and sort set values for stable exports.

            if isinstance(formatted_value, str):
                cell_value = formatted_value
            elif isinstance(formatted_value, Iterable):
                cell_value = ",".join(str(item) for item in formatted_value)
            else:

mmzeynalli · 2026-08-05T12:59:21Z

-                cell_value = ",".join(formatted_value)
+            if isinstance(formatted_value, str):
+                cell_value = formatted_value
+            elif isinstance(formatted_value, Iterable):


(list, set, tuple, frozenset)

Lets do this instead of Iterable. Apparently, formatted_value can be dict as well:

class Order(Base): items = relationship("Item", collection_class=attribute_keyed_dict("sku"))

Even, do like this:

if isinstance(value, bool): cell_value = "TRUE" if value else "FALSE" elif isinstance(formatted_value, str): cell_value = formatted_value elif isinstance(formatted_value, (set, frozenset)): cell_value = ",".join(sorted(str(item) for item in formatted_value)) elif isinstance(formatted_value, (list, tuple)): cell_value = ",".join(str(item) for item in formatted_value) else: cell_value = formatted_value

Bcuz, set and frozenset are unodered. and each export might give different set of items.

Applied your ladder in d4b0691, with the explicit types instead of Iterable.

if isinstance(formatted_value, str): cell_value = formatted_value elif isinstance(formatted_value, (set, frozenset)): cell_value = ",".join(sorted(str(item) for item in formatted_value)) elif isinstance(formatted_value, (list, tuple)): cell_value = ",".join(str(item) for item in formatted_value) else: cell_value = formatted_value

I kept the bool check in the non-relationship branch where it already was rather than hoisting it above the relationship branch - a bool can't be a relationship value, so moving it only changes which branch a non-relationship bool takes, and it currently takes that one.

Two tests added: test_get_export_row_values_with_set_relationship_is_sorted (three tags inserted as vip/beta/alpha, asserts alpha,beta,vip, so it pins the ordering rather than just the membership) and test_get_export_row_values_with_dict_relationship using collection_class=attribute_keyed_dict("sku"). Both fail on the previous commit - the dict one used to come out as sku-1 because Iterable caught the dict and joined its keys.

One thing to decide: with the explicit tuple, an attribute_keyed_dict relationship now falls through to else and the raw dict reaches the writer, so the CSV cell is the dict repr. That is what your snippet does and it is at least not silently wrong the way key-joining was, but if you'd rather it render as ",".join(sorted(str(v) for v in formatted_value.values())) say so and I'll add the branch plus flip the test.

Also changed the _base_export_cell return annotation from str to Any - it was never accurate, the else arms return whatever the formatter gave (int, date, None), and now a dict too.

mmzeynalli · 2026-09-03T15:24:46Z

Sign in to view

+            elif isinstance(formatted_value, Iterable):
+                cell_value = ",".join(str(item) for item in formatted_value)
            else:
                cell_value = formatted_value


Keyed collections (attribute_keyed_dict) aren't list/tuple/set/str, so they
fall through here and reach the CSV as a Python repr:

1,"Main St,Second St","alpha,vip","{'sku-1': <__main__.Item object at 0x7f55d58e03d0>, ...}"

Two runs of the same export over the same rows:

run 1: ...<__main__.Item object at 0x7f55d58e03d0>... run 2: ...<__main__.Item object at 0x7fe63fae0390>...

Which is the opposite of the comment three lines up — sets became reproducible in
the same commit that left mappings not reproducible. It also leaks the module path
and ignores __str__ on the related model, which every other branch honours.

9a3e5a6 had this right via Iterable (it produced sku-1,sku-2). Narrowing away
from Iterable was a sound instinct, but could we name dict rather than let it
fall through?

elif isinstance(formatted_value, dict): cell_value = ",".join(str(item) for item in formatted_value.values())

mmzeynalli · 2026-08-05T13:16:51Z

return type is Any

Copilot

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

Suppressed comments (1)

tests/test_pretty_export.py:190

This assertion assumes the addresses relationship is loaded in a deterministic order ("Main St" then "Second St"). Without an order_by on the relationship, most databases do not guarantee row ordering, so this test can be flaky depending on the backend/query plan. Since the intent is to verify stringification/joining, make the assertion order-insensitive (or add order_by to the relationship).

            assert len(values) == 3
            assert values[0] == 1
            assert values[1] == "John Doe"
            assert values[2] == "Main St,Second St"

mmzeynalli · 2026-09-03T15:24:46Z

+            elif isinstance(formatted_value, Iterable):
+                cell_value = ",".join(str(item) for item in formatted_value)
            else:
                cell_value = formatted_value


Keyed collections (attribute_keyed_dict) aren't list/tuple/set/str, so they
fall through here and reach the CSV as a Python repr:

1,"Main St,Second St","alpha,vip","{'sku-1': <__main__.Item object at 0x7f55d58e03d0>, ...}"

Two runs of the same export over the same rows:

run 1: ...<__main__.Item object at 0x7f55d58e03d0>... run 2: ...<__main__.Item object at 0x7fe63fae0390>...

Which is the opposite of the comment three lines up — sets became reproducible in
the same commit that left mappings not reproducible. It also leaks the module path
and ignores __str__ on the related model, which every other branch honours.

9a3e5a6 had this right via Iterable (it produced sku-1,sku-2). Narrowing away
from Iterable was a sound instinct, but could we name dict rather than let it
fall through?

elif isinstance(formatted_value, dict): cell_value = ",".join(str(item) for item in formatted_value.values())

mmzeynalli

Perfect. Almost ready! We just need one more test)

mmzeynalli · 2026-09-06T20:12:32Z

        else:
            if isinstance(value, bool):
                cell_value = "TRUE" if value else "FALSE"


Test missing for this line.

What still lands here is a to-one relationship, and on a related model without
__str__ that exports as the same thing this commit just fixed for mappings:

column_export_list = ["id", "author", "publisher"] Author has no __str__ -> <__main__.Author object at 0x7f...> Publisher has __str__ -> P CSV LINE -> 1,<__main__.Author object at 0x7f...>,P

Identical on current main, so not something this PR caused — but it's the same bug class,
so we might fix it in this PR

…elationships + test

Sanjays2402 · 2026-09-09T06:34:25Z

Thanks @mmzeynalli — that last test is in (ed1f827).

The to-one fall-through now honours __str__ like every other branch: cell_value = None if formatted_value is None else str(formatted_value), so a related model with __str__ exports its string form and a missing relation still exports as an empty cell (not the string "None").

Added test_get_export_row_values_with_to_one_relationship with Book/Publisher models — Publisher defines __str__ returning its name, so the test pins publisher exporting as "P", plus the None case. Also verified no line exceeds 88 chars for ruff.

Copilot AI lite review requested due to automatic review settings July 31, 2026 05:34

Copilot started reviewing on behalf of Sanjays2402 July 31, 2026 05:35 View session

Copilot AI reviewed Jul 31, 2026

View reviewed changes

mmzeynalli requested changes Aug 4, 2026

View reviewed changes

mmzeynalli added the needs-update There are stuff that needs updated and reviewed again label Aug 4, 2026

fix(pretty_export): handle set collections and string formatters for …

9a3e5a6

…relationships

Copilot AI review requested due to automatic review settings August 5, 2026 05:22

Copilot started reviewing on behalf of Sanjays2402 August 5, 2026 05:22 View session

Copilot AI reviewed Aug 5, 2026

View reviewed changes

mmzeynalli requested changes Aug 5, 2026

View reviewed changes

fix(pretty_export): sort set relationships and leave mappings untouched

d4b0691

Copilot AI review requested due to automatic review settings August 6, 2026 12:29

Copilot started reviewing on behalf of Sanjays2402 August 6, 2026 12:30 View session

Copilot AI reviewed Aug 6, 2026

View reviewed changes

mmzeynalli removed the needs-update There are stuff that needs updated and reviewed again label Sep 2, 2026

mmzeynalli requested changes Sep 3, 2026

View reviewed changes

mmzeynalli added the needs-update There are stuff that needs updated and reviewed again label Sep 3, 2026

fix: export keyed relationship values instead of object reprs

5586c9f

mmzeynalli requested changes Sep 6, 2026

View reviewed changes

mmzeynalli added waiting-for-tests Feature is ready, but tests are missing and removed needs-update There are stuff that needs updated and reviewed again labels Sep 8, 2026

Sanjays2402 added 2 commits September 8, 2026 23:34

Address review feedback from @mmzeynalli: honour __str__ for to-one r…

16f3fc3

…elationships + test

Fix line length (ruff E501) in to-one relationship branch

ed1f827

Fix to-one test: missing relation exports as empty cell

01a93b6

Uh oh!

Conversation

Sanjays2402 commented Jul 31, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mmzeynalli commented Aug 4, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

This comment was marked as resolved.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mmzeynalli left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Sanjays2402 commented Sep 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants