I'm trying to build a simple orm layer for golang.
Which would take a struct and generate the cols [] which can then be passed to sql function
rows.Scan(cols...) which takes pointers of fields in the struct corresponding to each of the columns it has found in the result set
Here is my example struct
type ExampleStruct struct {
ID int64 `sql:"id"`
aID string `sql:"a_id"`
UserID int64 `sql:"user_id"`
And this is my generic ORM function
func GetSqlColumnToFieldMap(model *ExampleStruct) map[string]interface{} {
typeOfModel := reflect.TypeOf(*model)
ValueOfModel := reflect.ValueOf(*model)
columnToDataPointerMap := make(map[string]interface{})
for i := 0; i < ValueOfModel.NumField(); i++ {
sql_column := typeOfModel.Field(i).Tag.Get("sql")
structValue := ValueOfModel.Field(i)
columnToDataPointerMap[sql_column] = structValue.Addr()
}
return columnToDataPointerMap
}
Once this method works fine i can use the map it generates to create an ordered list of sql pointers according to the column_names i get in rows() object
However i get below error on the .Addr() method call
panic: reflect.Value.Addr of unaddressable value [recovered]
panic: reflect.Value.Addr of unaddressable value
Is it not possible to do this ?
Also in an ideal scenario i would want the method to take an interface instead of *ExampleStruct so that it can be reused across different db models.
The error says the value whose address you want to get is unaddressable. This is because even though you pass a pointer to GetSqlColumnToFieldMap(), you immediately dereference it and work with a non-pointer value later on.
This value is wrapped in an interface{} when passed to reflect.ValueOf(), and values wrappped in interfaces are not addressable.
You must not dereference the pointer, but instead use Type.Elem() and Value.Elem() to get the element type and pointed value.
Something like this:
func GetSqlColumnToFieldMap(model *ExampleStruct) map[string]interface{} {
t := reflect.TypeOf(model).Elem()
v := reflect.ValueOf(model).Elem()
columnToDataPointerMap := make(map[string]interface{})
for i := 0; i < v.NumField(); i++ {
sql_column := t.Field(i).Tag.Get("sql")
structValue := v.Field(i)
columnToDataPointerMap[sql_column] = structValue.Addr()
}
return columnToDataPointerMap
}
With this simple change it works! And it doesn't depend on the parameter type, you may change it to interface{} and pass any struct pointers.
func GetSqlColumnToFieldMap(model interface{}) map[string]interface{} {
// ...
}
Testing it:
type ExampleStruct struct {
ID int64 `sql:"id"`
AID string `sql:"a_id"`
UserID int64 `sql:"user_id"`
}
type Point struct {
X int `sql:"x"`
Y int `sql:"y"`
}
func main() {
fmt.Println(GetSqlColumnToFieldMap(&ExampleStruct{}))
fmt.Println(GetSqlColumnToFieldMap(&Point{}))
}
Output (try it on the Go Playground):
map[a_id:<*string Value> id:<*int64 Value> user_id:<*int64 Value>]
map[x:<*int Value> y:<*int Value>]
Note that Value.Addr() returns the address wrapped in a reflect.Value. To "unwrap" the pointer, use Value.Interface():
func GetSqlColumnToFieldMap(model interface{}) map[string]interface{} {
t := reflect.TypeOf(model).Elem()
v := reflect.ValueOf(model).Elem()
m := make(map[string]interface{})
for i := 0; i < v.NumField(); i++ {
colName := t.Field(i).Tag.Get("sql")
field := v.Field(i)
m[colName] = field.Addr().Interface()
}
return m
}
This will output (try it on the Go Playground):
map[a_id:0xc00007e008 id:0xc00007e000 user_id:0xc00007e018]
map[x:0xc000018060 y:0xc000018068]
For an in-depth introduction to reflection, please read blog post: The Laws of Reflection
since few days I was struggling on how to proceed with PATCH request in Go REST API until I have found an article about using pointers and omitempty tag which I have populated and is working fine. Fine until I have realized I still have to build an UPDATE SQL query.
My struct looks like this:
type Resource struct {
Name *string `json:"name,omitempty" sql:"resource_id"`
Description *string `json:"description,omitempty" sql:"description"`
}
I am expecting a PATCH /resources/{resource-id} request containing such a request body:
{"description":"Some new description"}
In my handler I will build the Resource object this way (ignoring imports, ignoring error handling):
var resource Resource
resourceID, _ := mux.Vars(r)["resource-id"]
d := json.NewDecoder(r.Body)
d.Decode(&resource)
// at this point our resource object should only contain
// the Description field with the value from JSON in request body
Now, for normal UPDATE (PUT request) I would do this (simplified):
stmt, _ := db.Prepare(`UPDATE resources SET description = ?, name = ? WHERE resource_id = ?`)
res, _ := stmt.Exec(resource.Description, resource.Name, resourceID)
The problem with PATCH and omitempty tag is that the object might be missing multiple properties, thus I cannot just prepare a statement with hardcoded fields and placeholders... I will have to build it dynamically.
And here comes my question: how can I build such UPDATE query dynamically? In the best case I'd need some solution with identifying the set properties, getting their SQL field names (probably from the tags) and then I should be able to build the UPDATE query. I know I can use reflection to get the object properties but have no idea hot to get their sql tag name and of course I'd like to avoid using reflection here if possible... Or I could simply check for each property it is not nil, but in real life the structs are much bigger than provided example here...
Can somebody help me with this one? Did somebody already have to solve the same/similar situation?
SOLUTION:
Based on the answers here I was able to come up with this abstract solution. The SQLPatches method builds the SQLPatch struct from the given struct (so no concrete struct specific):
import (
"fmt"
"encoding/json"
"reflect"
"strings"
)
const tagname = "sql"
type SQLPatch struct {
Fields []string
Args []interface{}
}
func SQLPatches(resource interface{}) SQLPatch {
var sqlPatch SQLPatch
rType := reflect.TypeOf(resource)
rVal := reflect.ValueOf(resource)
n := rType.NumField()
sqlPatch.Fields = make([]string, 0, n)
sqlPatch.Args = make([]interface{}, 0, n)
for i := 0; i < n; i++ {
fType := rType.Field(i)
fVal := rVal.Field(i)
tag := fType.Tag.Get(tagname)
// skip nil properties (not going to be patched), skip unexported fields, skip fields to be skipped for SQL
if fVal.IsNil() || fType.PkgPath != "" || tag == "-" {
continue
}
// if no tag is set, use the field name
if tag == "" {
tag = fType.Name
}
// and make the tag lowercase in the end
tag = strings.ToLower(tag)
sqlPatch.Fields = append(sqlPatch.Fields, tag+" = ?")
var val reflect.Value
if fVal.Kind() == reflect.Ptr {
val = fVal.Elem()
} else {
val = fVal
}
switch val.Kind() {
case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
sqlPatch.Args = append(sqlPatch.Args, val.Int())
case reflect.String:
sqlPatch.Args = append(sqlPatch.Args, val.String())
case reflect.Bool:
if val.Bool() {
sqlPatch.Args = append(sqlPatch.Args, 1)
} else {
sqlPatch.Args = append(sqlPatch.Args, 0)
}
}
}
return sqlPatch
}
Then I can simply call it like this:
type Resource struct {
Description *string `json:"description,omitempty"`
Name *string `json:"name,omitempty"`
}
func main() {
var r Resource
json.Unmarshal([]byte(`{"description": "new description"}`), &r)
sqlPatch := SQLPatches(r)
data, _ := json.Marshal(sqlPatch)
fmt.Printf("%s\n", data)
}
You can check it at Go Playground. The only problem here I see is that I allocate both the slices with the amount of fields in the passed struct, which may be 10, even though I might only want to patch one property in the end resulting in allocating more memory than needed... Any idea how to avoid this?
I recently had same problem. about PATCH and looking around found this article. It also makes references to the RFC 5789 where it says:
The difference between the PUT and PATCH requests is reflected in the way the server processes the enclosed entity to modify the resource identified by the Request-URI. In a PUT request, the enclosed entity is considered to be a modified version of the resource stored on the origin server, and the client is requesting that the stored version be replaced. With PATCH, however, the enclosed entity contains a set of instructions describing how a resource currently residing on the origin server should be modified to produce a new version. The PATCH method affects the resource identified by the Request-URI, and it also MAY have side effects on other resources; i.e., new resources may be created, or existing ones modified, by the application of a PATCH.
e.g:
[
{ "op": "test", "path": "/a/b/c", "value": "foo" },
{ "op": "remove", "path": "/a/b/c" },
{ "op": "add", "path": "/a/b/c", "value": [ "foo", "bar" ] },
{ "op": "replace", "path": "/a/b/c", "value": 42 },
{ "op": "move", "from": "/a/b/c", "path": "/a/b/d" },
{ "op": "copy", "from": "/a/b/d", "path": "/a/b/e" }
]
This set of instructions should make it easier to build the update query.
EDIT
This is how you would obtain sql tags but you will have to use reflection:
type Resource struct {
Name *string `json:"name,omitempty" sql:"resource_id"`
Description *string `json:"description,omitempty" sql:"description"`
}
sp := "sort of string"
r := Resource{Description: &sp}
rt := reflect.TypeOf(r) // reflect.Type
rv := reflect.ValueOf(r) // reflect.Value
for i := 0; i < rv.NumField(); i++ { // Iterate over all the fields
if !rv.Field(i).IsNil() { // Check it is not nil
// Here you would do what you want to having the sql tag.
// Creating the query would be easy, however
// not sure you would execute the statement
fmt.Println(rt.Field(i).Tag.Get("sql")) // Output: description
}
}
I understand you don't want to use reflection, but still this may be a better answer than the previous one as you comment state.
EDIT 2:
About the allocation - read this guide lines of Effective Go about Data structures and Allocation:
// Here you are allocating an slice of 0 length with a capacity of n
sqlPatch.Fields = make([]string, 0, n)
sqlPatch.Args = make([]interface{}, 0, n)
With make(Type, Length, Capacity (optional))
Consider the following example:
// newly allocated zeroed value with Composite Literal
// length: 0
// capacity: 0
testSlice := []int{}
fmt.Println(len(testSlice), cap(testSlice)) // 0 0
fmt.Println(testSlice) // []
// newly allocated non zeroed value with make
// length: 0
// capacity: 10
testSlice = make([]int, 0, 10)
fmt.Println(len(testSlice), cap(testSlice)) // 0 10
fmt.Println(testSlice) // []
// newly allocated non zeroed value with make
// length: 2
// capacity: 4
testSlice = make([]int, 2, 4)
fmt.Println(len(testSlice), cap(testSlice)) // 2 4
fmt.Println(testSlice) // [0 0]
In your case, may want to the following:
// Replace this
sqlPatch.Fields = make([]string, 0, n)
sqlPatch.Args = make([]interface{}, 0, n)
// With this or simple omit the capacity in make above
sqlPatch.Fields = []string{}
sqlPatch.Args = []interface{}{}
// The allocation will go as follow: length - capacity
testSlice := []int{} // 0 - 0
testSlice = append(testSlice, 1) // 1 - 2
testSlice = append(testSlice, 1) // 2 - 2
testSlice = append(testSlice, 1) // 3 - 4
testSlice = append(testSlice, 1) // 4 - 4
testSlice = append(testSlice, 1) // 5 - 8
Alright, I think the solution I used back in 2016 was quite over-engineered for even more over-engineered problem and was completely unnecessary. The question asked here was very generalized, however we were building a solution that was able to build its SQL query on its own and based on the JSON object or query parameters and/or Headers sent in the request. And that to be as generic as possible.
Nowadays I think the best solution is to avoid PATCH unless truly necessary. And even then you still can use PUT and replace the whole resource with patched property/ies coming already from the client - i.e. not giving the client the option/possibility to send any PATCH request to your server and to deal with partial updates on their own.
However this is not always recommended, especially in cases of bigger objects to save some C02 by reducing the amount of redundant transmitted data. Whenever today if I need to enable a PATCH for the client I simply define what can be patched - this gives me clarity and the final struct.
Note that I am using a IETF documented JSON Merge Patch implementation. I consider that of JSON Patch (also documented by IETF) redundant as hypothetically we could replace the whole REST API by having one single JSON Patch endpoint and let clients control the resources via allowed operations. I also think the implementation of such JSON Patch on the server side is way more complicated. The only use-case I could think of using such implementation is if I was implementing a REST API over a file system...
So the struct may be defined as in my OP:
type ResourcePatch struct {
ResourceID some.UUID `json:"resource_id"`
Description *string `json:"description,omitempty"`
Name *string `json:"name,omitempty"`
}
In the handler func I'd decode the ID from the path into the ResourcePatch instance and unmarshall JSON from the request body into it, too.
Sending only this
{"description":"Some new description"}
to PATCH /resources/<UUID>
I should end up with with this object:
ResourcePatch
* ResourceID {"UUID"}
* Description {"Some new description"}
And now the magic: use a simple logic to build the query and exec parameters. For some it may seem tedious or repetitive or unclean for bigger PATCH objects, but my reply to this would be: if your PATCH object consists of more than 50% of the original resource' properties (or simply too many for your liking) use PUT and expect the clients to send (and replace) the whole resource instead.
It could look like this:
func (s Store) patchMyResource(r models.ResourcePatch) error {
q := `UPDATE resources SET `
qParts := make([]string, 0, 2)
args := make([]interface{}, 0, 2)
if r.Description != nil {
qParts = append(qParts, `description = ?`)
args = append(args, r.Description)
}
if r.Name != nil {
qParts = append(qParts, `name = ?`)
args = append(args, r.Name)
}
q += strings.Join(qParts, ',') + ` WHERE resource_id = ?`
args = append(args, r.ResourceID)
_, err := s.db.Exec(q, args...)
return err
}
I think there's nothing simpler and more effective. No reflection, no over-kills, reads quite good.
Struct tags are only visible through reflection, sorry.
If you don't want to use reflection (or, I think, even if you do), I think it is Go-like to define a function or method that "marshals" your struct into something that can easily be turned into a comma-separated list of SQL updates, and then use that. Build small things to help solve your problems.
For example given:
type Resource struct {
Name *string `json:"name,omitempty" sql:"resource_id"`
Description *string `json:"description,omitempty" sql:"description"`
}
You might define:
func (r Resource) SQLUpdates() SQLUpdates {
var s SQLUpdates
if (r.Name != nil) {
s.add("resource_id", *r.Name)
}
if (r.Description != nil) {
s.add("description", *r.Description)
}
}
where the type SQLUpdates looks something like this:
type SQLUpdates struct {
assignments []string
values []interface{}
}
func (s *SQLUpdates) add(key string, value interface{}) {
if (s.assignments == nil) {
s.assignments = make([]string, 0, 1)
}
if (s.values == nil) {
s.values = make([]interface{}, 0, 1)
}
s.assignments = append(s.assignments, fmt.Sprintf("%s = ?", key))
s.values = append(s.values, value)
}
func (s SQLUpdates) Assignments() string {
return strings.Join(s.assignments, ", ")
}
func (s SQLUpdates) Values() []interface{} {
return s.values
}
See it working (sorta) here: https://play.golang.org/p/IQAHgqfBRh
If you have deep structs-within-structs, it should be fairly easy to build on this. And if you change to an SQL engine that allows or encourages positional arguments like $1 instead of ?, it's easy to add that behavior to just the SQLUpdates struct without changing any code that used it.
For the purpose of getting arguments to pass to Exec, you would just expand the output of Values() with the ... operator.
I am attempting to run a conditional to basically see if the object is empty but I keep getting (similar variations) of this error:
invalid operation: release.Name == "" (mismatched types *string and string)
Here is the code that is dying:
import (
"github.com/google/go-github/github"
)
func TestLatestTag(user, project string) {
var client *github.Client
client = github.NewClient(nil)
releases, _, err := client.Repositories.ListTags(user, project, nil)
var release github.RepositoryTag
if err != nil {
fmt.Println("Error")
} else {
if release.Name == "" {
fmt.Println("None")
} else {
fmt.Println(releases[0])
}
}
}
If I change the if statement to *release.Name == "" as the error suggests I get a different error, which I don't really understand:
panic: runtime error: invalid memory address or nil pointer dereference
[signal 0xb code=0x1 addr=0x0 pc=0x26fd]
goroutine 1 [running]:
I'm sure there is any easy way to do this but I am not very familiar with handling objects/structs
From the error message it looks like you are trying to compare a string pointer (*string) to an actual string.
release.Name is a *string (a pointer to a string value)
"" is a string (is a string value)
They are two different types. So you can't compare them.
What you probably want to do instead is release.Name == nil
When a pointer that references to nothing (equals to nil) is tried to be dereferenced you get that second error. So in your case *release.Name panics because infact release.Name is nil
var release github.RepositoryTag
You never assign any value to that var. That's why *release.Name gives you a "runtime error": release.Name is a nil pointer
As per your code you have declared var release github.RepositoryTag, but you have not initialized it.
In structure RepositoryTag, Name is declared as *string which is a pointer and in case of release.Name == "", string comparison is attempted which is incorrect hence "mismatched types *string and string" error.
In case of *release.Name == "", since release is not yet initialized, it is complaining "invalid memory address or nil pointer dereference"
You need to do two things, 1st initialize, release and second, check release.Name = nil.
Go types like Int64 and String cannot store null values,
so I found I could use sql.NullInt64 and sql.NullString for this.
But when I use these in a Struct,
and generate JSON from the Struct with the json package,
then the format is different to when I use regular Int64 and String types.
The JSON has an additional level because the sql.Null*** is also a Struct.
Is there a good workaround for this,
or should I not use NULLs in my SQL database?
Types like sql.NullInt64 do not implement any special handling for JSON marshaling or unmarshaling, so the default rules apply. Since the type is a struct, it gets marshalled as an object with its fields as attributes.
One way to work around this is to create your own type that implements the json.Marshaller / json.Unmarshaler interfaces. By embedding the sql.NullInt64 type, we get the SQL methods for free. Something like this:
type JsonNullInt64 struct {
sql.NullInt64
}
func (v JsonNullInt64) MarshalJSON() ([]byte, error) {
if v.Valid {
return json.Marshal(v.Int64)
} else {
return json.Marshal(nil)
}
}
func (v *JsonNullInt64) UnmarshalJSON(data []byte) error {
// Unmarshalling into a pointer will let us detect null
var x *int64
if err := json.Unmarshal(data, &x); err != nil {
return err
}
if x != nil {
v.Valid = true
v.Int64 = *x
} else {
v.Valid = false
}
return nil
}
If you use this type in place of sql.NullInt64, it should be encoded as you expect.
You can test this example here: http://play.golang.org/p/zFESxLcd-c
If you use the null.v3 package, you won't need to implement any of the marshal or unmarshal methods. It's a superset of the sql.Null structs and is probably what you want.
package main
import "gopkg.in/guregu/null.v3"
type Person struct {
Name string `json:"id"`
Age int `json:"age"`
NickName null.String `json:"nickname"` // Optional
}
If you'd like to see a full Golang webserver that uses sqlite, nulls, and json you can consult this gist.
In my program, I store objective-c objects in a c array, like this
va_start(list, o);
retval->objs = malloc(SIZE * count);
retval->objs[0] = (__bridge void *)o;
for (int i = 1; i < count; i++)
{
id o = va_arg(list, id);
retval->objs[i] = (__bridge void *)o;
}
va_end(list);
(count is a number containing how many objects will be added; that value is always correct)
objs is a void ** and is part of retval, which is a pointer to a struct. As of now, SIZE is defined as 100. Increasing and decreasing that had no effect.
As you can see, I bridge o to a void *, as I have to. objs, when all the objects are added, contains 3 objective-c objects. When I try to access a value like this
void *obj = CLArrayObjectAtIndex(_arr, ind);
return (__bridge id)obj;
this is the CLArrayObjectAtIndex() function
void *CLArrayObjectAtIndex(CLArrayType *arr, int ind)
{
void *o = arr->objs[ind];
if (o)
return o;
else
perror("Attempt to access NULL object or index out of bounds."), abort();
}
if the index (ind) is 0, it works. If the index is 1, the program crashes when it returns in main. If the index is 2, the program crashes as soon as I try to access it. If the index is 1, the value returned above is correct, but when the program crashes on return it is nil.
If the index is 1, the EXC_BAD_ACCESS code is 1; if the index is 2, the code is EXC_I386_GPFLT, a general protection fault. I already checked here for an explanation of this exception, although I couldn't find anything helpful. So, does anybody see why this error may be occurring?
when you store obj-c objects in C array don't just bridge cast them since that way arc doesn't know they are still used and releases them. __bridge_retain them so they stay around later, when you free the array __bridge_transfer them to give them back to ARC
also don't define size as 100.. sizeof(id) should work. You only need to store pointers