Alert downstream maintainers of breaking changes

[krlmlr]

The reverse dependency failures across the R packages you listed share several recurring patterns that strongly suggest a common set of causes, primarily due to recent changes in the igraph package, particularly in version 2.1.5. Here’s a summary of the main patterns and root causes:

---

✅ 

Common Technical Patterns

1. 

Error in paste0() with type 'closure'

• Packages affected: alakazam, arulesViz, multivariance, others.

• Symptom:

Error in paste0(before, x, after) : cannot coerce type 'closure' to vector of type 'character'

• 
  

• Likely cause: A function name is being used instead of a variable due to namespace conflict or unintended variable masking. This is often triggered by how igraph or cli formats output with default attributes.

---

2. 

get_edge_ids() and get.edge.ids() 2x2 matrix error

• Packages affected: corpustools, dosearch, mwcsr, SEMID, klassR, MetaNet, manynet, others.

• Symptom:

Error: The `vp` argument of `get_edge_ids()` is not allowed to be a 2 times 2 matrix as of igraph 2.1.5.

• 
  

• Likely cause: Code previously passed a 2 x n matrix to igraph::get_edge_ids()—this is now deprecated in favor of n x 2.

• Solution: Transpose vp with t() or convert to a data.frame.

---

3. 

Coercion to dgCMatrix fails

• Packages affected: ccTensor, GMPro, manynet, SEMID, etc.

• Symptom:

Error in as(incidence, "dgCMatrix"): no method or default for coercing "matrix" to "dgCMatrix"

• 
  

• Likely cause: igraph::graph_from_biadjacency_matrix() expects a sparse matrix (Matrix::dgCMatrix), but is given a base matrix.

---

🧾 

Common Non-Fatal Notes Across Packages

• “Lost braces; missing escapes or markup?” in Rd files.

• Large installed size warnings (e.g., >5 MB libs).

• Non-ASCII characters in data or documentation.

These issues are generally not breaking builds, but may still be best addressed upstream.

---

🔄 Summary of Common Fixes Needed

Error Pattern	Fix
get_edge_ids() 2x2 matrix error	Use t(vp) or data.frame(from, to)
as(incidence, "dgCMatrix")	Convert to sparse matrix via Matrix::Matrix(incidence, sparse=TRUE)
paste0() closure error	Ensure variables are not masked by functions; check label, name, etc.
Rd formatting notes	Escape braces: { } or use \code{}
UTF-8 character notes	Validate encoding with tools::showNonASCIIfile()

---

📦 Commonly Affected Stack

• igraph (especially v2.1.5)

• cli, rlang, Matrix

• Plotting or graph construction packages using igraph, ggraph, or tidygraph

---

Would you like a script to help automatically flag and fix some of these patterns across multiple packages or a report template for notifying maintainers?

[krlmlr]

For 2x2: Users can use c() to convert to a vector or as.data.frame() .

The other issues: 🤷🏼‍♂️

[szhorvat]

• Likely cause: igraph::graph_from_biadjacency_matrix() expects a sparse matrix (Matrix::dgCMatrix), but is given a base matrix.

I'm confused: graph_from_biadjacency_matrix() must and does support dense matrices. What am I misunderstanding?

[szhorvat]

Was this summarized with an LLM? It just looks weird.

[krlmlr]

For the third issue, the as() will not work if the Matrix package is not loaded. I agree that a human should look at this. The machine-generated summary is but a hint on what to focus on first.

[szhorvat]

When posting machine-generated text, could you please add a short note at the top, just to make sure we don't get confused, and know what to expect? We also require this from people opening a PR. "This is an LLM-generated summary:" will do.

[schochastics]

#1797 should fix
Alakazam
arulesViz
discourseGT
multivariance
POSetR
remify
sglasso
simdata
wpa

[schochastics]

Scrolling through all failures I mostly noticed issues with hard deprecation of functions and the as() Matrix stuff

[krlmlr]

as() looks like we need to requireNamespace("Matrix", quietly = TRUE) before trying to coerce? Haven't looked in detail.

The hard deprecation is icky, we could tone down to a deprecate_warn() and make sure the error messages are as clear as we can make them.

[schochastics]

I agree with get_edge_ids(). We may have to rethink that, but there are some from 1.3 (and 1.6?) which is a long time to fix it:
The neimodeargument ofdfs() was deprecated in igraph 1.3.0

[krlmlr]

YOLO mode: fixed the sparse matrix issue in aa3cfc6, rerunning revdepchecks.

[schochastics]

I will propose a fix for get_edge_ids() later

[krlmlr]

I saw only one instance of that problem. With that, I'd rather break cleanly. Let's wait until the next round of results.

[schochastics]

Quick rundown:
ccTensor: as
corpusTools: get_edge_ids()
dosearch: no logical recycling
GMPro: as
Incidentally: ?
klassR: no logical recycling
manynet: as
MetaNet: as
mwcsr: get_edge_ids()
remify: new warning for NA in plotting
SEMID: get_edge_ids()
ssifs: neimode argument of dfs() was deprecated in igraph 1.3.0
tilemaps: The neimode argument of bfs() was deprecated in igraph 1.3.0