"""Regression tests for utils/pdf.py section + reference extraction.

Root causes fixed here: roman-numeral/plain-numbered headings ("VI. Results
and Discussion", "4.1 Methodology") collapsed the whole body into one section,
and references numbered "1. Author, Title." (no brackets) were never
extracted — which silently killed the citation-relevance agent and starved
the corpus-similarity pipeline of text.
"""

from __future__ import annotations

from utils.pdf import extract_references, split_into_sections


PAPER = """An AI-Driven Women Safety System
ABSTRACT
Women safety is a critical concern in modern society.
This paper presents an AI-driven women safety system.
I. Introduction
Women safety has become one of the most pressing social challenges.
II. Problem Statement and Motivation
Most existing women safety systems rely on manual SOS activation.
III. System Architecture
The proposed system follows a modular architecture.
IV. Methodology
The proposed methodology consists of voice-based distress detection.
4.1 Voice-Based Distress Detection
Audio input is captured in one-second intervals.
V. Implementation and Working
The proposed system is implemented as a mobile-based safety solution.
VI. Results and Discussion
The system demonstrated strong performance in detecting distress situations.
VII. Conclusion
This paper presented an AI-driven women safety system.
VIII. Future Scope
Expanding the training dataset can further improve model accuracy.
Acknowledgements
The authors would like to express their sincere gratitude.
REFERENCES
1. National Crime Records Bureau, Crime in India Report, Government of India, 2022.
2. S. Khan et al., "Deep Learning Techniques for Anomaly Detection," Elsevier, 2023.
3. Verma et al., "AI-Based Smart Safety Systems," International Journal of
Computer Applications, 2023.
"""


def test_section_split_handles_roman_numeral_headings():
    sections = split_into_sections(PAPER)
    keys = {k.strip().lower() for k in sections}
    assert "i. introduction" in keys
    assert "vi. results and discussion" in keys
    assert "viii. future scope" in keys
    assert "references" in keys


def test_section_split_keeps_body_out_of_abstract():
    sections = split_into_sections(PAPER)
    abstract = sections["abstract"]
    intro = next(v for k, v in sections.items() if "introduction" in k)
    assert "Women safety is a critical concern" in abstract
    assert "pressing social challenges" in intro
    assert "pressing social challenges" not in abstract


def test_section_split_dotted_number_heading():
    sections = split_into_sections(PAPER)
    meth = next(v for k, v in sections.items() if "methodology" in k)
    assert "one-second intervals" in meth


def test_extract_references_dot_numbering():
    refs = extract_references(PAPER.split("REFERENCES", 1)[1])
    assert len(refs) == 3
    assert "National Crime Records Bureau" in refs[0]
    assert "Deep Learning Techniques" in refs[1]
    assert "Computer Applications, 2023." in refs[2]


def test_extract_references_bracket_numbering_still_works():
    text = "[1] Alice Author, Title One. 2021.\n[2] Bob Author, Title Two. 2022."
    refs = extract_references(text)
    assert len(refs) == 2
    assert refs[0].startswith("Alice Author")


def test_extract_references_parenthesized_numbering():
    text = "(1) First Reference Here. 2020.\n(2) Second Reference There. 2021."
    refs = extract_references(text)
    assert len(refs) == 2
    assert refs[1].startswith("Second Reference")


def test_extract_references_empty_and_garbage():
    assert extract_references("") == []
    assert extract_references("Some plain paragraph without numbers.") == []