HOTFIX: Fix rapi_bind length mismatch when saving citation networks

[seanthimons]

Hotfix

Issue: When saving a citation network: Error saving network: rapi_bind: Bind parameter values need to have the same length

Root cause: Two issues in save_network() (R/db.R):

1. source_notebook_id = NULL passed directly to dbExecute as a zero-length bind parameter — DuckDB requires all bind params to have equal length
2. nodes_df$community %||% NA_character_ and nodes_df$is_overlap %||% FALSE produce scalar (length-1) fallbacks when columns are absent, mixed with length-N columns in the data.frame passed to dbWriteTable

Fix:

1. Coerce NULL → NA_character_ for source_notebook_id before binding
2. Replace scalar %||% fallbacks with rep() to ensure length-N vectors

Triage findings:

• Code trace: Both dbExecute (metadata insert) and dbWriteTable (nodes bulk insert) had parameter length mismatches
• Git history: The community column addition and missing NULL handling were introduced at different times
• Test status: Existing tests always passed source_notebook_id = "nb-1" and included is_overlap/community columns, masking the bug

Test plan

• Existing test suite passes (no new failures introduced)
• New regression test: NULL source_notebook_id + missing is_overlap/community columns
• Manual verification: save a network from Citation Network tab (no notebook context)

[Copilot]

Pull request overview

Hotfix to prevent DuckDB bind-parameter length mismatches when saving citation networks, especially in “standalone” contexts where source_notebook_id is NULL and some node columns aren’t present.

Changes:

• Coerces source_notebook_id from NULL to NA_character_ before dbExecute() to avoid zero-length bind parameters.
• Ensures is_overlap / community fallbacks are length-N vectors (via rep(...)) to avoid dbWriteTable() column length mismatches.
• Adds regression tests covering NULL source_notebook_id and missing is_overlap/community.

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 2 comments.

File	Description
R/db.R	Fixes parameter binding and data.frame column length handling in save_network()
tests/testthat/test-save-network.R	Adds regression coverage for the failing save scenario and a “full columns present” case

---

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

[Copilot]

save_network() always includes a community column in the data.frame passed to dbWriteTable(…, "network_nodes", …, append = TRUE). In the repo migrations, network_nodes is created in migrations/006_create_citation_networks.sql without community, and migrations/010_add_multi_seed_support.sql adds is_overlap but not community. A fresh DB built from migrations will therefore likely fail when saving a network. Please add a migration to ALTER TABLE network_nodes ADD COLUMN IF NOT EXISTS community VARCHAR (and consider making save_network() resilient by omitting community when the column doesn’t exist).

[Copilot]

The test creates a custom network_nodes schema that includes community and is_overlap. This diverges from the schema managed by repo migrations (e.g., migrations/006_create_citation_networks.sql doesn’t define community and migrations/010_add_multi_seed_support.sql doesn’t add it), so the regression test can pass even when a fresh migrated DB would fail. Consider initializing the schema via the project’s migration/init helpers (or otherwise deriving the test table definitions from the migration SQL) so the test stays aligned with production schema.

[seanthimons]

Recreating branch from integration to resolve conflicts