Compare · 004

Nullability through joins, COALESCE and aggregates: DuckDB and ClickHouse

Every file of the DuckDB and ClickHouse versions side by side, with the lines that differ marked. 3 parameter or column types differ in Go.

Compare DuckDB withSQLitePostgreSQLCockroachDBTimescaleDBMySQLMariaDBClickHouse

Configuration

sgl.yamldiffers
DuckDB
ClickHouse
1version: 1
1version: 1
2
2
3databases:
3databases:
4 - name: colony
4 - name: colony
5 dialect: duckdb
5 dialect: clickhouse
6 migrations: db/migrations
6 migrations: db/migrations
7 queries: db/queries
7 queries: db/queries
8 go:
8 go:
9 package: store
9 package: store
10 out: internal/store
10 out: internal/store

Migrations

db/migrations/0001_gulls.up.sqldiffers
DuckDB
ClickHouse
1CREATE SEQUENCE gull_ids;
2
3CREATE SEQUENCE sighting_ids;
4
5CREATE TABLE gulls (
1CREATE TABLE gulls (
6 id BIGINT PRIMARY KEY DEFAULT nextval('gull_ids'),
2 id UInt64,
7 ring VARCHAR NOT NULL UNIQUE,
3 ring String,
8 nickname VARCHAR
4 nickname Nullable(String)
9);
5) ENGINE = MergeTree() ORDER BY id;
10
6
11CREATE TABLE sightings (
7CREATE TABLE sightings (
12 id BIGINT PRIMARY KEY DEFAULT nextval('sighting_ids'),
8 id UInt64,
13 gull_id BIGINT NOT NULL REFERENCES gulls (id),
9 gull_id UInt64,
14 seen_on DATE NOT NULL,
10 seen_on Date,
15 place VARCHAR NOT NULL
11 place String
16);
12) ENGINE = MergeTree() ORDER BY id;
db/migrations/0001_gulls.down.sqldiffers
DuckDB
ClickHouse
1DROP TABLE sightings;
1DROP TABLE sightings;
2DROP TABLE gulls;
2DROP TABLE gulls;
3DROP SEQUENCE sighting_ids;
4DROP SEQUENCE gull_ids;

Queries

RingGulldiffers
DuckDB · db/queries/gulls.sql
ClickHouse · db/queries/gulls.sql
1-- sgl.query(RingGull, exec)
1-- sgl.query(RingGull, exec)
2INSERT INTO gulls (ring, nickname) VALUES ($1, $2);
2INSERT INTO gulls (id, ring, nickname) VALUES ({id:UInt64}, {ring:String}, {nickname:Nullable(String)});
RecordSightingdiffers
DuckDB · db/queries/gulls.sql
ClickHouse · db/queries/gulls.sql
1-- sgl.query(RecordSighting, exec)
1-- sgl.query(RecordSighting, exec)
2INSERT INTO sightings (gull_id, seen_on, place) VALUES ($1, $2, $3);
2INSERT INTO sightings (id, gull_id, seen_on, place) VALUES ({id:UInt64}, {gull_id:UInt64}, {seen_on:Date}, {place:String});
ListSightingsidentical
DuckDB · db/queries/gulls.sql
ClickHouse · db/queries/gulls.sql
1-- sgl.query(ListSightings, many)
1-- sgl.query(ListSightings, many)
2SELECT gulls.ring, gulls.nickname, sightings.place
2SELECT gulls.ring, gulls.nickname, sightings.place
3FROM gulls
3FROM gulls
4LEFT JOIN sightings ON sightings.gull_id = gulls.id
4LEFT JOIN sightings ON sightings.gull_id = gulls.id
5ORDER BY gulls.id, sightings.seen_on;
5ORDER BY gulls.id, sightings.seen_on;
SummariseGullsdiffers
DuckDB · db/queries/gulls.sql
ClickHouse · db/queries/gulls.sql
1-- sgl.query(SummariseGulls, many)
1-- sgl.query(SummariseGulls, many)
2SELECT
2SELECT
3 gulls.ring,
3 ring,
4 COALESCE(gulls.nickname, 'ring ' || gulls.ring) AS known_as,
4 coalesce(nickname, concat('ring ', ring)) AS known_as,
5 COUNT(sightings.id) AS times_seen,
5 count(seen_on) AS times_seen,
6 MAX(sightings.seen_on) AS last_seen
6 max(seen_on) AS last_seen
7FROM gulls
7FROM gulls
8LEFT JOIN sightings ON sightings.gull_id = gulls.id
8LEFT JOIN sightings ON sightings.gull_id = gulls.id
9GROUP BY gulls.id, gulls.ring, gulls.nickname
9GROUP BY gulls.id, ring, nickname
10ORDER BY gulls.id;
10ORDER BY gulls.id
11SETTINGS join_use_nulls = 1;

What sgl generates

RingGulldiffers
DuckDB
ClickHouse
1func (queries *Queries) RingGull(ctx context.Context, params RingGullParams) error
1func (queries *Queries) RingGull(ctx context.Context, params RingGullParams) error
2
2
3type RingGullParams struct {
3type RingGullParams struct {
4 ID uint64
4 Ring string
5 Ring string
5 Nickname *string
6 Nickname *string
6}
7}
RecordSightingdiffers
DuckDB
ClickHouse
1func (queries *Queries) RecordSighting(ctx context.Context, params RecordSightingParams) error
1func (queries *Queries) RecordSighting(ctx context.Context, params RecordSightingParams) error
2
2
3type RecordSightingParams struct {
3type RecordSightingParams struct {
4 GullID int64
4 ID uint64
5 GullID uint64
5 SeenOn time.Time
6 SeenOn time.Time
6 Place string
7 Place string
7}
8}
ListSightingsdiffers
DuckDB
ClickHouse
1func (queries *Queries) ListSightings(ctx context.Context) ([]ListSightingsRow, error)
1func (queries *Queries) ListSightings(ctx context.Context) ([]ListSightingsRow, error)
2
2
3type ListSightingsRow struct {
3type ListSightingsRow struct {
4 Ring string `json:"ring"`
4 Ring string `json:"ring"`
5 Nickname *string `json:"nickname"`
5 Nickname *string `json:"nickname"`
6 Place *string `json:"place"`
6 Place string `json:"place"`
7}
7}
SummariseGullsdiffers
DuckDB
ClickHouse
1func (queries *Queries) SummariseGulls(ctx context.Context) ([]SummariseGullsRow, error)
1func (queries *Queries) SummariseGulls(ctx context.Context) ([]SummariseGullsRow, error)
2
2
3type SummariseGullsRow struct {
3type SummariseGullsRow struct {
4 Ring string `json:"ring"`
4 Ring string `json:"ring"`
5 KnownAs string `json:"known_as"`
5 KnownAs string `json:"known_as"`
6 TimesSeen int64 `json:"times_seen"`
6 TimesSeen uint64 `json:"times_seen"`
7 LastSeen *time.Time `json:"last_seen"`
7 LastSeen *time.Time `json:"last_seen"`
8}
8}

The program

example.godiffers
DuckDB
ClickHouse
1package main
1package main
2
2
3import (
3import (
4 "context"
4 "context"
5 "database/sql"
5 "database/sql"
6 "fmt"
6 "fmt"
7 "io"
7 "io"
8 "time"
8 "time"
9
9
10 "seagull.sh/seagull/examples/duckdb/004_nullability/internal/store"
10 "seagull.sh/seagull/examples/clickhouse/004_nullability/internal/store"
11)
11)
12
12
13// run rings three gulls, records where two of them were seen, and lists them twice.
13// run rings three gulls, records where two of them were seen, and lists them twice.
14// Seagull decides each field's nullability from the SQL, and a field that can be NULL is
14// Seagull decides each field's nullability from the SQL, and a field that can be NULL is
15// a pointer:
15// a pointer:
16//
16//
17// - gulls.nickname has no NOT NULL, so Nickname is *string, as a parameter and a
17// - gulls.nickname is Nullable(String), so Nickname is *string, as a parameter and a
18// column;
18// column; a ClickHouse column is never NULL unless its type is Nullable(T);
19// - sightings.place is NOT NULL, but the LEFT JOIN gives a gull with no sightings a
19// - ListSightings leaves join_use_nulls at the server's default, 0, so its LEFT JOIN
20// NULL place, so Place is *string;
20// fills an unmatched row with each column's default value instead of NULL. Place is a
21// - COALESCE with a non-NULL last argument is never NULL, so KnownAs is string;
21// String column, so it stays a string, and Gwen, who was never seen, gets an empty
22// - COUNT is never NULL, but MAX over no rows is, so TimesSeen is int64 and LastSeen is
22// one;
23// *time.Time, as seen_on is a DATE.
23// - SummariseGulls sets join_use_nulls = 1, so the unmatched row is NULL: count skips
24// it and gives 0, and max over no dates gives NULL, so TimesSeen is uint64 and
25// LastSeen is *time.Time. Without the setting, count would see Gwen's default row and
26// give 1;
27// - coalesce with a fallback that is never NULL is never NULL, so KnownAs is string.
24//
28//
25// Takes database (*sql.DB) which is a migrated database.
29// Takes database (*sql.DB) which is a migrated database.
26// Takes output (io.Writer) which receives both listings.
30// Takes output (io.Writer) which receives both listings.
27//
31//
28// Returns error when a query fails.
32// Returns error when a query fails.
29func run(ctx context.Context, database *sql.DB, output io.Writer) error {
33func run(ctx context.Context, database *sql.DB, output io.Writer) error {
30 queries := store.New(database)
34 queries := store.New(database)
31
35
32 gulls := []store.RingGullParams{
36 gulls := []store.RingGullParams{
33 {Ring: "GB1001", Nickname: new("Steven")},
37 {ID: 1, Ring: "GB1001", Nickname: new("Steven")},
34 {Ring: "GB1002", Nickname: nil},
38 {ID: 2, Ring: "GB1002", Nickname: nil},
35 {Ring: "GB1003", Nickname: new("Gwen")},
39 {ID: 3, Ring: "GB1003", Nickname: new("Gwen")},
36 }
40 }
37 for _, gull := range gulls {
41 for _, gull := range gulls {
38 if err := queries.RingGull(ctx, gull); err != nil {
42 if err := queries.RingGull(ctx, gull); err != nil {
39 return fmt.Errorf("ringing %s: %w", gull.Ring, err)
43 return fmt.Errorf("ringing %s: %w", gull.Ring, err)
40 }
44 }
41 }
45 }
42 sightings := []store.RecordSightingParams{
46 sightings := []store.RecordSightingParams{
43 {GullID: 1, SeenOn: day(1), Place: "the pier"},
47 {ID: 1, GullID: 1, SeenOn: day(1), Place: "the pier"},
44 {GullID: 1, SeenOn: day(3), Place: "the chip shop"},
48 {ID: 2, GullID: 1, SeenOn: day(3), Place: "the chip shop"},
45 {GullID: 2, SeenOn: day(2), Place: "the harbour wall"},
49 {ID: 3, GullID: 2, SeenOn: day(2), Place: "the harbour wall"},
46 }
50 }
47 for _, sighting := range sightings {
51 for _, sighting := range sightings {
48 if err := queries.RecordSighting(ctx, sighting); err != nil {
52 if err := queries.RecordSighting(ctx, sighting); err != nil {
49 return fmt.Errorf("recording a sighting: %w", err)
53 return fmt.Errorf("recording a sighting: %w", err)
50 }
54 }
51 }
55 }
52
56
53 rows, err := queries.ListSightings(ctx)
57 rows, err := queries.ListSightings(ctx)
54 if err != nil {
58 if err != nil {
55 return fmt.Errorf("listing sightings: %w", err)
59 return fmt.Errorf("listing sightings: %w", err)
56 }
60 }
57 fmt.Fprintln(output, "ring nickname place")
61 fmt.Fprintln(output, "ring nickname place")
58 for _, row := range rows {
62 for _, row := range rows {
59 fmt.Fprintf(output, "%-7s %-9s %s\n", row.Ring, orNull(row.Nickname), orNull(row.Place))
63 fmt.Fprintf(output, "%-7s %-9s %s\n", row.Ring, orNull(row.Nickname), orEmpty(row.Place))
60 }
64 }
61
65
62 summaries, err := queries.SummariseGulls(ctx)
66 summaries, err := queries.SummariseGulls(ctx)
63 if err != nil {
67 if err != nil {
64 return fmt.Errorf("summarising gulls: %w", err)
68 return fmt.Errorf("summarising gulls: %w", err)
65 }
69 }
66 fmt.Fprintln(output, "\nring known as times seen last seen")
70 fmt.Fprintln(output, "\nring known as times seen last seen")
67 for _, summary := range summaries {
71 for _, summary := range summaries {
68 lastSeen := "NULL"
72 lastSeen := "NULL"
69 if summary.LastSeen != nil {
73 if summary.LastSeen != nil {
70 lastSeen = summary.LastSeen.Format(time.DateOnly)
74 lastSeen = summary.LastSeen.Format(time.DateOnly)
71 }
75 }
72 fmt.Fprintf(output, "%-7s %-12s %-11d %s\n", summary.Ring, summary.KnownAs, summary.TimesSeen, lastSeen)
76 fmt.Fprintf(output, "%-7s %-12s %-11d %s\n", summary.Ring, summary.KnownAs, summary.TimesSeen, lastSeen)
73 }
77 }
74 return nil
78 return nil
75}
79}
76
80
77// day returns a date in May 2026, the month the sightings were made.
81// day returns a date in May 2026, at midnight UTC. A Date parameter binds the calendar
82// date a time has in its own location.
78//
83//
79// Takes dayOfMonth (int) which is the day.
84// Takes date (int) which is the day of the month.
80//
85//
81// Returns time.Time which is midnight UTC on that day.
86// Returns time.Time which is the date.
82func day(dayOfMonth int) time.Time {
87func day(date int) time.Time {
83 return time.Date(2026, time.May, dayOfMonth, 0, 0, 0, 0, time.UTC)
88 return time.Date(2026, time.May, date, 0, 0, 0, 0, time.UTC)
84}
89}
85
90
86// orNull returns the text a nullable field holds, or NULL when the field is nil.
91// orNull returns the text a nullable field holds, or NULL when the field is nil.
87//
92//
88// Takes value (*string) which is a nullable field.
93// Takes value (*string) which is a nullable field.
89//
94//
90// Returns string which is the text to print.
95// Returns string which is the text to print.
91func orNull(value *string) string {
96func orNull(value *string) string {
92 if value == nil {
97 if value == nil {
93 return "NULL"
98 return "NULL"
94 }
99 }
95 return *value
100 return *value
96}
101}
102
103// orEmpty returns a string field's text, or (empty) when it holds the empty string, the
104// default value ClickHouse gives a String column in an unmatched outer-join row.
105//
106// Takes value (string) which is a String column.
107//
108// Returns string which is the text to print.
109func orEmpty(value string) string {
110 if value == "" {
111 return "(empty)"
112 }
113 return value
114}

Go type differences

The same SQL types map to different Go types, because each database reports them differently.

QueryFieldDuckDBClickHouse
RecordSightingparameter gull_idint64uint64
ListSightingscolumn place*stringstring
SummariseGullscolumn times_seenint64uint64