Type Mismatch Error with Multiple Character Fields

[marcosnav]

Summary

get_content() fails with type mismatch errors when the Connect server returns content items where some have character fields (like cluster_name, r_version, py_version, title, etc.) as character and others have them as logical/NULL. This affects multiple optional character fields, not just one.

Error Messages

Various errors depending on which field has the type mismatch:

Error in dplyr::bind_rows: Can't combine `..1$cluster_name` <character> and `..3$cluster_name` <logical>.
Error in dplyr::bind_rows: Can't combine `..1$r_version` <character> and `..2$r_version` <logical>.
Error in dplyr::bind_rows: Can't combine `..1$title` <character> and `..88$title` <logical>.

The pattern is consistent: character fields that can be NULL for certain content types are being parsed inconsistently.

Environment

• connectapi version: 0.8.0 (also affects 0.10.0)
• Connect version: 2026.03.0-dev+6-g34e1d74381
• R version: 4.5.0
• dplyr version: 1.1.4
• purrr version: 1.0.2

Root Cause

The issue occurs in the parse_connectapi() function when processing content data:

1. The Connect API returns content items where:

   • Content in clusters has: cluster_name = "cluster-name-string"
   • Content not in clusters has: cluster_name = null

2. When R processes these:

   • Character values remain character type
   • NULL values become logical NA

3. parse_connectapi() uses purrr::map_df() which calls dplyr::bind_rows()

4. bind_rows() fails immediately when trying to combine character and logical columns

5. The ptype definition (cluster_name = NA_character_) exists but is applied after parsing via ensure_columns(), so it never gets a chance to fix the types

Impact

This affects any application using get_content() when connected to a Connect server that has:

• Cluster support enabled
• A mix of content items in and out of clusters

Reproducibility

This should be reproducible by:

1. Setting up a Connect server with cluster support
2. Creating some content in a cluster and some outside
3. Calling get_content() via connectapi

Additional Context

The ptype.R file already defines all these fields with their correct types in the content prototype (e.g., cluster_name = NA_character_, r_version = NA_character_, etc.), showing the library authors intended these fields to be character type. The issue is that this normalization happens too late in the process - after bind_rows() has already failed.

Fields known to be affected:

• cluster_name - NULL for content not in clusters
• r_version - NULL for non-R content
• py_version - NULL for non-Python content
• quarto_version - NULL for non-Quarto content
• image_name - NULL for non-containerized content
• title - NULL for certain content items
• Potentially any other optional character field in the content schema

[karawoo]

This is fixed in the latest connectapi release (0.11.0) via #503. I'm going to close this issue but please reopen if you continue to have problems with 0.11.0.