https://github.com/mwblythe/squint
An interpolating SQL builder for Go
https://github.com/mwblythe/squint
database go golang sql
Last synced: about 1 month ago
JSON representation
An interpolating SQL builder for Go
- Host: GitHub
- URL: https://github.com/mwblythe/squint
- Owner: mwblythe
- License: mit
- Created: 2020-06-08T04:21:57.000Z (almost 6 years ago)
- Default Branch: master
- Last Pushed: 2025-12-02T21:06:26.000Z (4 months ago)
- Last Synced: 2025-12-05T21:32:53.905Z (4 months ago)
- Topics: database, go, golang, sql
- Language: Go
- Homepage:
- Size: 82 KB
- Stars: 3
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Squint - An interpolating SQL builder
## Overview
The `database/sql` package is powerful but somewhat tedious to use. You must hand-write full SQL queries with bind placeholders and provide a matching ordered list of variables. It's familiar, but inconvenient and repetitive. Squint makes things easier by allowing SQL and bind variables to be intermixed in their natural order. It also interpolates the variables into the proper bind placeholders and values, including complex types like structs and maps. Squint is not an ORM, though. It's merely a pleasant query building assistant.
## Builder
The `Builder` is the heart of Squint. It accepts a list of SQL fragments and variables and returns the resulting SQL statement and bind variables.
```go
import "github.com/mwblythe/squint"
// a simple example
idList := []int{10, 20, 30}
b := squint.NewBuilder()
sql, binds := b.Build("select * from users where id in", idList, "and active = 1")
rows, err := db.Query(sql, binds...)
```
The `sql` and `binds` returned are ready to pass into the standard `database/sql` query functions, such as `Exec`, `Query` and `QueryRow`.
💡 See [driver](driver) for an easier way.
### Basic Variables
A `string` by default is treated as SQL. Other variables of the [basic types](https://github.com/golang/example/tree/master/gotypes#basic-types) will transform into a SQL bind placeholder (`?`) and an accompanying bind value. To have a `string` treated as a bind variable, you can pass a reference (`&myString`) or cast it as `squint.Bind(myString)`.
```go
// treated as binds
b.Build("select * from users where id =", 10) // good
b.Build("select * from users where is_active =", true) // good
// strings are sql without special handling
name := "Frank"
b.Build("select * from users where name =", name) // bad
b.Build("select * from users where name =", &name) // good
b.Build("select * from users where name =", squint.Bind(name)) // good
```
### Arrays and Slices
By default, arrays and slices will be flattened and treated as if their contents had been passed inline. However, as part of a SQL `IN` clause, they will be expanded into appropriate placeholders and binds (even `strings`).
```go
bits := []string{"select *", "from", "crew"}
b.Build(bits, "where id =", 10) // slice is flattened
names := []string{"jim", "spock", "uhura"}
b.Build("select * from crew where name in", names) // magic
```
*Note that only arrays of basic types are supported for `IN` clause.*
### Structs and Maps
By default, these will be expanded in the style of a `WHERE` clause (`column = ?`) and joined with `AND`.
```go
// easily build where clauses from structs
type Query struct {
Origin string
Completed bool
}
q := Query{Origin: "online", Completed: false}
b.Build("select * from orders where", q)
// or maps
type M map[string]interface{}
q := M{"origin": "online", "completed": false }
b.Build("select * from orders where", q)
```
There is special handling for `INSERT` statements:
```go
// simple user struct
type User struct {
Id int
Name string
}
// fill structure, likely from user input
newUser := User{10, "Frank"}
// build our query
b.Build("insert into users", newUser)
// multi-row insert
users := []User{
{ 10, "Frank"},
{ 20, "Hank"},
}
b.Build("insert into users", users)
// also handles maps
type Item map[string]interface{}
b.Build("insert into users", Item{"id": 10, "name": "Frank"})
```
Also for `UPDATE` statements:
```go
// use a structure
type Updates struct {
Role string
Department string
}
updates := Updates{"captain", "housewares"}
b.Build("UPDATE user SET", updates, "WHERE id =", id)
// or a map
type Updates map[string]interface{}
updates := Updates{"role": "captain", "department": "housewares"}
b.Build("UPDATE user SET", updates, "WHERE id =", id)
```
### Pointers
Generally, pointers are dereferenced and their values used as if they were passed directly. If the pointer is `nil`, it will map to a `NULL` value. Pointers can be useful in a `struct` as discussed below under "Empty Values".
### Conditions
When crafting a complex query, you sometimes need to build it up in bits while checking various conditions. Was an ID specified? Was extra information requested? While you can do this by carefully filling an array that you then pass to `Build()`, Squint has another option.
```go
b.Build(
"SELECT u.* FROM users u",
b.If(boolCondition, "JOIN employees e ON u.id = e.id"),
"WHERE u.id IN", ids,
)
```
You can include any number of arguments in `If()`, and they will only be processed by `Build()` if the condition is true. This can also be called as `squint.If()`
### Field Mapping
When mapping `struct` fields into database columns, by default the names are used verbatim. You can change the mapping by using the `db` struct field.
```go
type User struct {
Id int // column is Id
FirstName string `db:"first_name"` // column is first_name
Username string `db:"-"` // skip this one
ManagerId int `db:"mgr_id,omitempty"` // skip if empty (0)
}
```
*A custom mapping function may be a future enhancement*
### Options
The `Builder` uses functional options to control behavior:
| Option | Purpose | Default |
| ----------------------------- | ------------------------------------------------------- | ------- |
| `Tag(string)` | tag name for field mapping | "db" |
| `KeepEmpty()` | keep empty values in struct/map | On |
| `OmitEmpty()` | omit empty values in struct/map | Off |
| `NullEmpty()` | treat empty values as nulls in struct/map | Off |
| `WithEmptyFn(squint.EmptyFn)` | use a custom empty value handler | nil |
| `WithDefaultEmpty()` | use default empty value handler | On |
| `LogQuery(bool)` | log queries | `false` |
| `LogBinds(bool)` | log bind values | `false` |
| `Log(bool)` | shorthand to log both queries AND binds | `false` |
| `BindQuestion()` | use `?` as bind placeholders (mysql, sqlite) | On |
| `BindDollar()` | use `$1, $2` style bind placeholders (postgres, sqlite) | Off |
| `BindAt()` | use `@p1, @p2` style placeholders (sqlserver) | Off |
| `BindColon()` | use `:b1, :b2` style placeholders (oracle) | Off |
| `WithBindFn(squint.BindFn)` | use a custom bind placeholder function | Off |
These can all be set via `NewBuilder()`:
```
b := squint.NewBuilder(
squint.NullEmpty(),
squint.Log(true),
)
```
They can also be set via `Build()` and will only be in effect for that query:
```go
b.Build(
squint.LogBinds(false),
"update users set password =", &myPass,
"where id =", id,
)
```
### Empty Values
When a struct or map is processed, empty (Go "zero") values need special consideration. You can control how they are treated on a builder level with the `KeepEmpty()`, `OmitEmpty()`, and `NullEmpty()` options. These are mutually exclusive, so only the last one used will win. Each of these options has a struct field equivalent for selective override:
```go
type Person struct {
Name string
Age int `db:"omitempty"`
NumChildren int `db:"keepempty"`
ManagerID int `db:"nullempty"`
}
```
Since the empty value for pointers is `nil`, you can sometimes leverage this:
```go
type Updates struct {
Department string
Balance *float64 `db:"omitempty"`
}
```
Now you can tell the difference between setting `Balance` to `0` or not setting it at all. With `omitempty`, an empty `Balance` would be skipped, but a pointer to a `0` would be kept.
**NOTE:** If `OmitEmpty()` is in effect for a multi-row inserts, `KeepEmpty()` will be used instead. This is because the column count must be consistent across rows. `NullEmpty()` and `KeepEmpty()` will be used as set.
### Custom Empty Handler
Squint can also use your custom handler functions for empty values. When an empty value is encountered, your function will be called instead of the default logic. This means that any custom function will override the behavior of the `KeepEmpty()`, `OmitEmpty()`, and `NullEmpty()` options. Any field-level tags will still be respected, however.
A custom function looks like this:
```go
func(in interface{}) (out interface{}, keep bool)
```
The parameter `in` is the empty value in question. Your function should return `keep` which determines if the empty value should be kept, and `out` which is the value to use if kept.
You register it via the `squint.WithEmptyFn()` option:
```go
// use "N/A" for empty strings, skip all other types
func doEmpty(in interface{}) (out interface{}, keep bool) {
if s, ok := in.(string); ok {
return "N/A", true
}
return nil, false
}
// set globally in Builder
b := squint.NewBuilder(
squint.WithEmptyFn(doEmpty)
)
// override for single query
b.Build(
squint.WithEmptyFn(otherEmpty),
"insert into users", newUser,
)
```
To switch back to the default empty value handler, use the `WithDefaultEmpty()` option.
## See Also
For a more seamless solution, see the squint [driver](driver) package.