feat: error when adding named vertices to non-empty unnamed graphs

[Copilot]

Adding named vertices to unnamed graphs creates confusing output where existing vertices have NA names and edges display as NA--NA.

Changes

• Error on name mismatch: Added error (abort) in +.igraph when:
  • Using vertex()/vertices() with unnamed arguments (interpreted as names) on non-empty unnamed graphs (graphs with vertices)
  • Using character vectors (e.g., g + c("a", "b")) on non-empty unnamed graphs
• Allow empty graphs: Adding named vertices to empty graphs (graphs with 0 vertices, created by make_empty_graph()) is allowed, enabling graph construction from scratch
• Test coverage: Added snapshot tests for error messages in tests/testthat/_snaps/operators.md
• Test updates: Updated tests to work with the new error behavior

The implementation checks vcount(e1) > 0 to determine if a graph is non-empty. Empty graphs are defined as having zero vertices, consistent with standard graph theory terminology.

Example

# Error when adding to non-empty unnamed graphs
> make_ring(10) + vertex(1)
Error: Cannot add named vertices to a non-empty unnamed graph. Existing vertices will have `NA` names.

> make_empty_graph(1) + vertex("a")
Error: Cannot add named vertices to a non-empty unnamed graph. Existing vertices will have `NA` names.

# Works when adding to empty graphs (0 vertices)
> make_empty_graph() + vertex("a")
IGRAPH DN-- 1 0 -- 
+ attr: name (v/c)

> make_empty_graph() + c("a", "b")
IGRAPH UN-- 2 0 -- 
+ attr: name (v/c)

# Works when adding to already-named graphs
> g <- make_ring(10); V(g)$name <- letters[1:10]; g + vertex("k")
IGRAPH UN-- 11 10 -- Ring graph
+ edges from ... (vertex names):
 [1] a--b b--c c--d ...

# Works when adding unnamed vertices (attributes only)
> make_ring(10) + vertex(foo = 5)
IGRAPH U--- 11 10 -- Ring graph
+ attr: name (g/c), mutual (g/l), circular (g/l), foo (v/n)

• Fixes Warn when adding named vertices to unnamed graphs #60

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Warn when adding named vertices to unnamed graphs</issue_title>
<issue_description>Or add vertex names automatically? Or both? E.g.

make_ring(10) + vertex(1)
#> IGRAPH UN-- 11 10 -- Ring graph
#> + attr: name (g/c), mutual (g/l), circular (g/l), name (v/n)
#> + edges (vertex names):
#>  [1] NA--NA NA--NA NA--NA NA--NA NA--NA NA--NA NA--NA NA--NA NA--NA NA--NA

</issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes Warn when adding named vertices to unnamed graphs #60

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

[krlmlr]

@copilot: Abort instead of warning. Adding to an empty graph should work. Remember to add a snapshot test for the error message.

[Copilot]

@copilot: Abort instead of warning. Adding to an empty graph should work. Remember to add a snapshot test for the error message.

Changed to abort instead of warn in commit 8723eee. The implementation now:

• Errors when adding named vertices to non-empty unnamed graphs (graphs with edges)
• Allows adding named vertices to empty graphs (no edges), like make_empty_graph() or make_empty_graph(n)
• Added snapshot tests for the error messages in tests/testthat/_snaps/operators.md

The key insight is checking ecount(e1) > 0 rather than vertex count, which prevents the confusing NA--NA edge display issue while still allowing graph construction from scratch.

[krlmlr]

@copilot: Empty graphs have zero vertices, not edges. Review the error message.

[Copilot]

@copilot: Empty graphs have zero vertices, not edges. Review the error message.

Corrected in commit a4aa43d. Changed the check from ecount(e1) > 0 to vcount(e1) > 0. Empty graphs are now properly defined as having 0 vertices. The error message is now consistent with the check - "non-empty" refers to vertex count, not edge count.

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 533b865 is merged into main:

• ✔️as_adjacency_matrix: 797ms -> 799ms [-0.47%, +1.1%]
• ✔️as_biadjacency_matrix: 774ms -> 772ms [-1.13%, +0.46%]
• ✔️as_data_frame_both: 1.63ms -> 1.62ms [-2.34%, +1.49%]
• ✔️as_long_data_frame: 4.17ms -> 4.2ms [-0.98%, +2.51%]
• ✔️es_attr_filter: 2.91ms -> 2.89ms [-2.53%, +1.57%]
• ✔️graph_from_adjacency_matrix: 119ms -> 119ms [-0.64%, +1.1%]
• ✔️graph_from_data_frame: 3.55ms -> 3.56ms [-0.34%, +1.25%]
• ✔️vs_attr_filter: 1.65ms -> 1.63ms [-2.88%, +0.74%]
• ✔️vs_by_name: 1.06ms -> 1.06ms [-2.19%, +2.95%]
  Further explanation regarding interpretation and methodology can be found in the documentation.