Fixture files that start with <% (erb) insert doc inside the erb block

[kwerle]

Steps to reproduce the problem

My yaml fixture

<%
define some hash
loop_times do
%>
some:
    block: of stuff
<%
end
%>
etc...

annotaterb models --force

Expected behavior

The schema information should NOT be placed in the erb block.

Actual behavior

The schema information goes into the erb block.

<%
# == Schema Information
# ...
define some hash
loop_times do
%>
some:
    block: of stuff
<%
end
%>
...

Your .annotaterb.yml configuration

---
active_admin: false
additional_file_patterns: []
routes: false
models: true
position_in_routes: before_doc
position_in_class: before_doc
position_in_test: before_doc
position_in_fixture: before_doc
position_in_factory: before_doc
position_in_serializer: before_doc
show_foreign_keys: true
show_complete_foreign_keys: false
show_indexes: true
simple_indexes: false
model_dir:
  - app/models
root_dir:
  - ""
include_version: false
require: []
exclude_tests: false
exclude_fixtures: false
exclude_factories: false
exclude_serializers: false
exclude_scaffolds: true
exclude_controllers: true
exclude_helpers: true
exclude_sti_subclasses: false
ignore_model_sub_dir: false
ignore_columns:
ignore_routes:
ignore_unknown_models: false
hide_limit_column_types: integer,bigint,boolean
hide_default_column_types: json,jsonb,hstore
skip_on_db_migrate: false
format_bare: true
format_rdoc: false
format_yard: false
format_markdown: false
sort: false
force: false
frozen: <%= !ENV["TESTING_ANNOTATED_MODELS"].to_s.empty? %>
classified_sort: true
trace: false
wrapper_open:
wrapper_close:
with_comment: true

Environment

• AnnotateRb version: github main branch
• Ruby version: 4.0
• Ruby on Rails version: 8.0
• Database adapter (e.g., sqlite3, pg, mysql2): mysql2 0.5.6

[Halvanhelv]

Opened #353 to fix this.

Root cause: ERB fixtures can't be parsed by Psych, so the parser evaluated the ERB and read Psych's line numbers from the evaluated output, then used them as indices into the original file. When an ERB tag spans multiple lines the offsets drift, so the annotation gets written inside the ERB block. (It also affected the <% ... do %> loop case — the schema ended up inside the loop body, duplicating it on every iteration.)

The PR stops evaluating the ERB and derives the content bounds straight from the original lines, so annotations land above the ERB body and never inside a tag.