summaryrefslogtreecommitdiffstats
path: root/modules/indexer/issues/internal/model.go
blob: 2dfee8b72e197bbb173773b06e6672626997a46e (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
// Copyright 2023 The Gitea Authors. All rights reserved.
// SPDX-License-Identifier: MIT

package internal

import (
	"code.gitea.io/gitea/models/db"
	"code.gitea.io/gitea/modules/optional"
	"code.gitea.io/gitea/modules/timeutil"
)

// IndexerData data stored in the issue indexer
type IndexerData struct {
	ID       int64 `json:"id"`
	RepoID   int64 `json:"repo_id"`
	IsPublic bool  `json:"is_public"` // If the repo is public

	// Fields used for keyword searching
	Title    string   `json:"title"`
	Content  string   `json:"content"`
	Comments []string `json:"comments"`

	// Fields used for filtering
	IsPull             bool               `json:"is_pull"`
	IsClosed           bool               `json:"is_closed"`
	LabelIDs           []int64            `json:"label_ids"`
	NoLabel            bool               `json:"no_label"` // True if LabelIDs is empty
	MilestoneID        int64              `json:"milestone_id"`
	ProjectID          int64              `json:"project_id"`
	ProjectColumnID    int64              `json:"project_board_id"` // the key should be kept as project_board_id to keep compatible
	PosterID           int64              `json:"poster_id"`
	AssigneeID         int64              `json:"assignee_id"`
	MentionIDs         []int64            `json:"mention_ids"`
	ReviewedIDs        []int64            `json:"reviewed_ids"`
	ReviewRequestedIDs []int64            `json:"review_requested_ids"`
	SubscriberIDs      []int64            `json:"subscriber_ids"`
	UpdatedUnix        timeutil.TimeStamp `json:"updated_unix"`

	// Fields used for sorting
	// UpdatedUnix is both used for filtering and sorting.
	// ID is used for sorting too, to make the sorting stable.
	CreatedUnix  timeutil.TimeStamp `json:"created_unix"`
	DeadlineUnix timeutil.TimeStamp `json:"deadline_unix"`
	CommentCount int64              `json:"comment_count"`
}

// Match represents on search result
type Match struct {
	ID    int64   `json:"id"`
	Score float64 `json:"score"`
}

// SearchResult represents search results
type SearchResult struct {
	Total int64
	Hits  []Match
}

// SearchOptions represents search options.
//
// It has a slightly different design from database query options.
// In database query options, a field is never a pointer, so it could be confusing when it's zero value:
// Do you want to find data with a field value of 0, or do you not specify the field in the options?
// To avoid this confusion, db introduced db.NoConditionID(-1).
// So zero value means the field is not specified in the search options, and db.NoConditionID means "== 0" or "id NOT IN (SELECT id FROM ...)"
// It's still not ideal, it trapped developers many times.
// And sometimes -1 could be a valid value, like issue ID, negative numbers indicate exclusion.
// Since db.NoConditionID is for "db" (the package name is db), it makes sense not to use it in the indexer:
// Why do bleve/elasticsearch/meilisearch indexers need to know about db.NoConditionID?
// So in SearchOptions, we use pointer for fields which could be not specified,
// and always use the value to filter if it's not nil, even if it's zero or negative.
// It can handle almost all cases, if there is an exception, we can add a new field, like NoLabelOnly.
// Unfortunately, we still use db for the indexer and have to convert between db.NoConditionID and nil for legacy reasons.
type SearchOptions struct {
	Keyword string // keyword to search

	IsFuzzyKeyword bool // if false the levenshtein distance is 0

	RepoIDs   []int64 // repository IDs which the issues belong to
	AllPublic bool    // if include all public repositories

	IsPull   optional.Option[bool] // if the issues is a pull request
	IsClosed optional.Option[bool] // if the issues is closed

	IncludedLabelIDs    []int64 // labels the issues have
	ExcludedLabelIDs    []int64 // labels the issues don't have
	IncludedAnyLabelIDs []int64 // labels the issues have at least one. It will be ignored if IncludedLabelIDs is not empty. It's an uncommon filter, but it has been supported accidentally by issues.IssuesOptions.IncludedLabelNames.
	NoLabelOnly         bool    // if the issues have no label, if true, IncludedLabelIDs and ExcludedLabelIDs, IncludedAnyLabelIDs will be ignored

	MilestoneIDs []int64 // milestones the issues have

	ProjectID       optional.Option[int64] // project the issues belong to
	ProjectColumnID optional.Option[int64] // project column the issues belong to

	PosterID optional.Option[int64] // poster of the issues

	AssigneeID optional.Option[int64] // assignee of the issues, zero means no assignee

	MentionID optional.Option[int64] // mentioned user of the issues

	ReviewedID        optional.Option[int64] // reviewer of the issues
	ReviewRequestedID optional.Option[int64] // requested reviewer of the issues

	SubscriberID optional.Option[int64] // subscriber of the issues

	UpdatedAfterUnix  optional.Option[int64]
	UpdatedBeforeUnix optional.Option[int64]

	Paginator *db.ListOptions

	SortBy SortBy // sort by field
}

// Copy returns a copy of the options.
// Be careful, it's not a deep copy, so `SearchOptions.RepoIDs = {...}` is OK while `SearchOptions.RepoIDs[0] = ...` is not.
func (o *SearchOptions) Copy(edit ...func(options *SearchOptions)) *SearchOptions {
	if o == nil {
		return nil
	}
	v := *o
	for _, e := range edit {
		e(&v)
	}
	return &v
}

type SortBy string

const (
	SortByCreatedDesc  SortBy = "-created_unix"
	SortByUpdatedDesc  SortBy = "-updated_unix"
	SortByCommentsDesc SortBy = "-comment_count"
	SortByDeadlineDesc SortBy = "-deadline_unix"
	SortByCreatedAsc   SortBy = "created_unix"
	SortByUpdatedAsc   SortBy = "updated_unix"
	SortByCommentsAsc  SortBy = "comment_count"
	SortByDeadlineAsc  SortBy = "deadline_unix"
	// Unsupported sort types which are supported by issues.IssuesOptions.SortType:
	//
	//  - "priorityrepo":
	//                    It's impossible to support it in the indexer.
	//                    It is based on the specified repository in the request, so we cannot add static field to the indexer.
	//                    If we do something like that query the issues in the specified repository first then append other issues,
	//                    it will break the pagination.
	//
	// - "project-column-sorting":
	//                    Although it's possible to support it by adding project.ProjectIssue.Sorting to the indexer,
	//                    but what if the issue belongs to multiple projects?
	//                    Since it's unsupported to search issues with keyword in project page, we don't need to support it.
)