package obs
import (
"regexp"
"time"
)
// A swarm is a crawler that sends each request from a different address
// with a browser's name: residential proxies, hundreds of networks and
// a hundred countries in one minute, one request per address, almost
// all of them for pages named by a hash (commit diffs, check-ins). Each
// request alone looks like a person, so the traffic page counted them
// as people. The sweep keeps what such requests share, their network
// ranges and how they presented themselves, for longer than the
// addresses, so patterns can be found and the ranges handled later.
// SwarmKept is how long the swarm's rows are kept: the hourly counts'
// window, long enough to see a pattern recur.
const SwarmKept = 90 * 24 * time.Hour
// SwarmPage is a path the swarm is counted on: one that names a version
// by its hash (what a reader reaches by following a link), or one of
// the pages that read a repository's whole history and so cost the
// forge the most (on 2026-11-10 three swarm requests for a large
// mirror's activity page took it down).
var SwarmPage = regexp.MustCompile(`/[0-9a-f]{10,64}(/(diff|patch|manifest|tags))?$` +
`|^/r/(?:[^/?]+/){1,2}(?:activity|timeline|search|history(?:/[^?]*)?|archive/[^?]*)(?:\?.*)?$`)
// hashInPath is any hash in a path, for its shape.
var hashInPath = regexp.MustCompile(`/[0-9a-f]{10,64}`)
// PathShape is a path with its hashes taken out, so requests for
// different versions of the same kind of page count together.
func PathShape(path string) string { return hashInPath.ReplaceAllString(path, "/{hash}") }
// SwarmCount is one value and how often it was seen.
type SwarmCount struct {
Key string `json:"key"`
Count int64 `json:"count"`
}
// SwarmDay is one day (or hour) of the swarm.
type SwarmDay struct {
At time.Time `json:"at"`
Requests int64 `json:"requests"`
Prefixes int64 `json:"prefixes"`
Networks int64 `json:"networks"`
}
// SwarmPrefix is one network range the swarm came from.
type SwarmPrefix struct {
Prefix string `json:"prefix"`
ASN uint32 `json:"asn"`
Network string `json:"network"`
Country string `json:"country"`
Requests int64 `json:"requests"`
Days int64 `json:"days"`
First time.Time `json:"first"`
Last time.Time `json:"last"`
}
// SwarmNetwork is one network the swarm came from.
type SwarmNetwork struct {
ASN uint32 `json:"asn"`
Network string `json:"network"`
Requests int64 `json:"requests"`
Prefixes int64 `json:"prefixes"`
}
// Swarm is what the sweep has kept, read back.
type Swarm struct {
From time.Time `json:"from"`
To time.Time `json:"to"`
Requests int64 `json:"requests"`
Days []SwarmDay `json:"days"`
Hours []SwarmDay `json:"hours"`
Prefixes []SwarmPrefix `json:"prefixes"`
Networks []SwarmNetwork `json:"networks"`
Hints []SwarmCount `json:"hints"`
Langs []SwarmCount `json:"langs"`
HeadSets []SwarmCount `json:"headSets"`
Shapes []SwarmCount `json:"shapes"`
Repos []SwarmCount `json:"repos"`
Agents []SwarmCount `json:"agents"`
Countries []SwarmCount `json:"countries"`
}
// Swarm reads the swarm's rows for the days given. A forge with history
// switched off has none, and says so with a nil.
func (r *Recorder) Swarm(days int) (*Swarm, error) {
r.mu.Lock()
sink := r.sink
r.mu.Unlock()
s, ok := sink.(interface {
Swarm(now time.Time, days int) (*Swarm, error)
})
if !ok {
return nil, nil
}
return s.Swarm(time.Now(), days)
}