from __future__ import annotations

import copy

import numpy as np
import pytest

from pallet.reclassify_appearance_outliers import align_features, classify_features, enrich_rows


def test_centers_are_sample_weighted_and_never_use_outliers():
    x=np.array([[0.,0.],[0.,0.],[0.,0.],[8.,0.],[20.,0.],[22.,0.],[999.,0.]])
    labels=np.array([1,1,1,1,2,2,-1])
    result=classify_features(x,labels)
    np.testing.assert_array_equal(result['centers'],[[2.,0.],[21.,0.]])
    np.testing.assert_array_equal(result['reference_counts'],[4,2])
    np.testing.assert_array_equal(result['centers'],classify_features(x[:-1],labels[:-1])['centers'])
    assert result['final_labels'][-1]==2 and result['far'][-1] and result['priority_review'][-1]


def test_exact_distance_tie_uses_smaller_class_and_zero_margin():
    result=classify_features([[0.],[2.],[9.],[11.],[5.5]],np.array([4,4,9,9,-1]))
    assert result['final_labels'][-1]==4
    assert result['nearest_distance'][-1]==4.5 and result['second_distance'][-1]==4.5
    assert result['margin'][-1]==0 and result['ambiguous'][-1]


def test_reference_labels_retained_when_their_nearest_center_disagrees():
    x=np.array([[0.],[0.],[9.],[10.],[10.],[8.]])
    labels=np.array([1,1,1,2,2,-1]);result=classify_features(x,labels)
    assert result['nearest_class'][2]==2
    assert result['final_labels'][2]==1 and not result['reassigned'][2]
    assert result['final_labels'][-1]==2


def test_cached_feature_indices_aligned_by_identity_even_if_merged_rows_reordered():
    source=[{'target_id':name,'appearance_cluster':i} for i,name in enumerate(['a','b','c'])]
    merged=[source[i]|{'merged_class_id':1} for i in [2,0,1]]
    # Cached row order is b,c,a; the merged file order is c,a,b.
    result=align_features(merged,source,np.array([1,2,0]),np.array([[20.],[30.],[10.]]))
    np.testing.assert_array_equal(result,[[30.],[10.],[20.]])
    with pytest.raises(ValueError,match='Duplicate|duplicate'):
        align_features(merged,source,np.array([1,1,0]),np.array([[20.],[30.],[10.]]))
    with pytest.raises(ValueError,match='missing'):
        align_features(merged,source,np.array([1,0]),np.array([[20.],[10.]]))


def test_changed_original_target_metadata_rejected():
    source=[{'target_id':'a','bbox':{'x':1.5}}]
    with pytest.raises(ValueError,match='metadata changed'):
        align_features([{'target_id':'a','bbox':{'x':2.}}],source,np.array([0]),np.array([[1.]]))


def test_enriched_rows_preserve_all_previous_fields_and_mark_every_new_label_unreviewed():
    x=np.array([[0.],[2.],[10.],[12.],[1.],[100.]])
    labels=np.array([1,1,2,2,-1,-1]);result=classify_features(x,labels)
    rows=[{'target_id':str(i),'domain':'a','merged_class_id':int(c) if c>0 else None,
           'merged_class':f'Class {c}' if c>0 else 'unassigned','appearance_membership':.5 if c>0 else 0.,
           'bbox':{'x':1.5,'y':2.25}} for i,c in enumerate(labels)]
    before=copy.deepcopy(rows)
    groups={1:{'merged_class':'Class 1','domains':{'a':4}},2:{'merged_class':'Class 2','domains':{'b':4}}}
    output=enrich_rows(rows,result,groups)
    assert rows==before
    assert all(all(r[k]==v for k,v in original.items()) for original,r in zip(rows,output))
    assert output[4]['classification_review_required'] and not output[4]['classification_priority_review']
    assert output[5]['classification_review_required'] and output[5]['classification_priority_review']
    assert not output[5]['assigned_class_domain_supported']
    assert output[5]['merged_class']=='unassigned' and output[5]['final_class']=='Class 2'
    assert output[0]['classification_source']=='retained_excel_merge'


def test_coincident_centers_are_ambiguous_and_zero_radius_never_serializes_infinity():
    x=np.array([[0.],[0.],[0.],[0.],[0.],[1.]])
    labels=np.array([1,1,2,2,-1,-1]);result=classify_features(x,labels)
    rows=[{'target_id':str(i),'domain':'a'} for i in range(len(x))]
    output=enrich_rows(rows,result,{i:{'merged_class':f'Class {i}','domains':{}} for i in [1,2]})
    assert output[4]['centroid_top2_margin']==0 and output[4]['distance_to_reference_p95_ratio']==0
    assert output[5]['distance_to_reference_p95_ratio'] is None
    assert output[5]['classification_review_flags']==['beyond_reference_p95','small_top2_margin']
    import json
    json.dumps(output,allow_nan=False)


@pytest.mark.parametrize('features,labels',[
    ([[0.],[float('nan')],[2.]],[1,2,-1]),
    ([[0.],[float('inf')],[2.]],[1,2,-1]),
    ([[0.],[1.]],[1,1]),
    ([[0.],[1.]],[0,2]),
    ([[0.],[1.]],[1,-2]),
    ([[0.],[1.]],[1]),
])
def test_invalid_or_incomplete_feature_spaces_rejected(features,labels):
    with pytest.raises(ValueError):classify_features(features,np.array(labels))


def test_assignments_and_centers_repeat_exactly():
    rng=np.random.default_rng(42);x=rng.normal(size=(150,34));labels=np.array([1]*40+[2]*60+[-1]*50)
    a,b=classify_features(x,labels),classify_features(x,labels)
    for key in a:np.testing.assert_array_equal(a[key],b[key])
