[BUG] Jinja: re.search not truthy when regex search found

[boxysean]

Describe the bug
Similar to #560, a successful re.search does not return truthy value when successful.

What version of dbt Fusion is this bug in? (find out by running dbt --version)
dbt-fusion 2.0.0-preview.2

Is this a discrepancy between the dbt Fusion Engine and dbt Core? Check one.

• YES
• NO

To Reproduce

1. Open this workspace
2. Follow the instructions in the README

Or these original instructions:

{%- set re = modules.re -%}
{%- set re_dur = re.compile('.*') -%}
{%- set dur = re.search(re_dur, 'xyz') -%}

{% if dur %}
select {{ dur }} as should_be_true
{% endif %}

-- Core compiles to: "select <re.Match object; span=(0, 3), match='xyz'> as should_be_true"
-- Fusion compiles to: <empty>

Expected behavior
Fusion's modules.re.search returns truthy like Core does

Screenshots
N/A

Operating System and CPU Type (please complete the following information):
Mac Apple Silicon

[dbeatty10]

Thanks for opening this @boxysean !

I got the same results with dbt-core 1.10.9 and dbt-fusion 2.0.0-preview.3 -- did you run this with dbt-core and different results than Fusion? If so, could you share the code you used and the output in dbt-core vs. Fusion so we can verify that there is a discrepancy?

Observed behavior across Core and Fusion

A re.search() operation does not return a truthy value when a match is found.

Reprex

Create these files:

models/my_model.sql

{%- set pat = modules.re.compile('.*') -%}
{%- set res = pat.search('abc') -%}
select {{ res == true }} as should_be_true

Run these commands using dbt-core:

dbt compile --select my_model

Get this output:

Compiled node 'my_model' is:
select False as should_be_true

Run these commands with Fusion:

dbt compile --select my_model

Get this output:

select False as should_be_true

More testing

I used the Claude Code CLI to compare / contrast a few different scenarios below:

Test 1: Original Issue Pattern (models/my_model.sql)

{%- set pat = modules.re.compile('.*') -%}
{%- set res = pat.search('abc') -%}
select {{ res == true }} as should_be_true

dbt-core result: select False as should_be_true
dbt-fusion result: select False as should_be_true

Test 2: Detailed Regex Analysis (models/regex_test.sql)

{%- set pat = modules.re.compile('.*') -%}
{%- set res = pat.search('abc') -%}
{%- set res_none = pat.search('') -%}

select 
    {{ res == true }} as res_eq_true,
    {{ res }} as res_raw,
    {{ res is not none }} as res_is_not_none,
    {{ res_none == true }} as res_none_eq_true,
    {{ res_none }} as res_none_raw,
    {{ res_none is not none }} as res_none_is_not_none

Both engines produce:

select 
    False as res_eq_true,
    <re.Match object; span=(0, 3), match='abc'> as res_raw,
    True as res_is_not_none,
    False as res_none_eq_true,
    <re.Match object; span=(0, 0), match=''> as res_none_raw,
    True as res_none_is_not_none

Test 3: Truthiness Check (models/truthiness_test.sql)

{%- set pat = modules.re.compile('.*') -%}
{%- set res = pat.search('abc') -%}
{%- set res_none = pat.search('') -%}

select 
    {% if res %}True{% else %}False{% endif %} as res_truthy_if,
    {% if res_none %}True{% else %}False{% endif %} as res_none_truthy_if

Both engines produce:

select 
    True as res_truthy_if,
    True as res_none_truthy_if

Test 4: Combined Analysis (models/original_issue.sql)

{%- set pat = modules.re.compile('.*') -%}
{%- set res = pat.search('abc') -%}
{%- set no_match = modules.re.search('xyz', 'abc') -%}

select 
    {{ res == true }} as should_be_true,
    {% if res %}True{% else %}False{% endif %} as res_is_truthy,
    {{ no_match == true }} as no_match_should_be_false,
    {% if no_match %}True{% else %}False{% endif %} as no_match_is_truthy

Both engines produce:

select 
    False as should_be_true,
    True as res_is_truthy,
    False as no_match_should_be_false,
    False as no_match_is_truthy

[boxysean]

Ah you are right @dbeatty10. I've updated the repro, can you check again?

[dbeatty10]

@boxysean It looks like I was looking at the wrong place for the compiled output for Fusion earlier in this thread. The key insight is that Core and Fusion place compiled files in different locations by default:

• dbt-core: target/compiled/[PROJECT_NAME]/models/[MODEL_NAME].sql
• dbt-fusion: target/compiled/models/[MODEL_NAME].sql

👉 Core adds PROJECT_NAME to the compilation path in the target directory whereas Fusion does not.

Differences observed

I re-ran all the scenarios above plus the update repro you provided -- see below for the output.

Quick summary of differences observed:

1. Boolean casing differences: Fusion mixes false (lowercase) and True (Title Case) inconsistently
2. Information loss: Fusion renders Match objects as empty {} instead of showing the actual match details
3. Compilation failures: The user's example still produces empty files in Fusion

Reprex

Used Claude Code to run the following scenarios and capture the output:

Test 1: Original Issue Pattern (models/my_model.sql)

{%- set pat = modules.re.compile('.*') -%}
{%- set res = pat.search('abc') -%}
select {{ res == true }} as should_be_true

dbt-core result: select False as should_be_true
dbt-fusion result: select false as should_be_true (lowercase)

Test 2: Detailed Regex Analysis (models/regex_test.sql)

{%- set pat = modules.re.compile('.*') -%}
{%- set res = pat.search('abc') -%}
{%- set res_none = pat.search('') -%}

select 
    {{ res == true }} as res_eq_true,
    {{ res }} as res_raw,
    {{ res is not none }} as res_is_not_none,
    {{ res_none == true }} as res_none_eq_true,
    {{ res_none }} as res_none_raw,
    {{ res_none is not none }} as res_none_is_not_none

dbt-core result:

select 
    False as res_eq_true,
    <re.Match object; span=(0, 3), match='abc'> as res_raw,
    True as res_is_not_none,
    False as res_none_eq_true,
    <re.Match object; span=(0, 0), match=''> as res_none_raw,
    True as res_none_is_not_none

dbt-fusion result:

select 
    false as res_eq_true,
    {} as res_raw,
    true as res_is_not_none,
    false as res_none_eq_true,
    {} as res_none_raw,
    true as res_none_is_not_none

Test 3: Truthiness Check (models/truthiness_test.sql)

{%- set pat = modules.re.compile('.*') -%}
{%- set res = pat.search('abc') -%}
{%- set res_none = pat.search('') -%}

select 
    {% if res %}True{% else %}False{% endif %} as res_truthy_if,
    {% if res_none %}True{% else %}False{% endif %} as res_none_truthy_if

Both engines produce identical results:

select 
    True as res_truthy_if,
    True as res_none_truthy_if

Test 4: Combined Analysis (models/original_issue.sql)

{%- set pat = modules.re.compile('.*') -%}
{%- set res = pat.search('abc') -%}
{%- set no_match = modules.re.search('xyz', 'abc') -%}

select 
    {{ res == true }} as should_be_true,
    {% if res %}True{% else %}False{% endif %} as res_is_truthy,
    {{ no_match == true }} as no_match_should_be_false,
    {% if no_match %}True{% else %}False{% endif %} as no_match_is_truthy

dbt-core result:

select 
    False as should_be_true,
    True as res_is_truthy,
    False as no_match_should_be_false,
    False as no_match_is_truthy

dbt-fusion result:

select 
    false as should_be_true,
    True as res_is_truthy,
    false as no_match_should_be_false,
    False as no_match_is_truthy

Test 5: User-Provided Example

{%- set re = modules.re -%}
{%- set re_dur = re.compile('.*') -%}
{%- set dur = re.search(re_dur, 'xyz') -%}

{% if dur %}
select {{ dur }} as should_be_true
{% endif %}

dbt-core result (fresh target):

select <re.Match object; span=(0, 3), match='xyz'> as should_be_true

dbt-fusion result (fresh target):

• EMPTY FILE (0 bytes) - Complete compilation failure