I'm trying to make a simple tool which parses JSON-formatted lines in a file and performs an INSERT operation into a database.
I have a struct which looks like this:
type DataBlob struct {
....
Datetime time.Time `json:"datetime, string"`
....
}
And parsing code which looks like this:
scanner := bufio.NewScanner(file)
// Loop through all lines in the file
for scanner.Scan() {
var t DataBlob
// Decode the line, parse the JSON
dec := json.NewDecoder(strings.NewReader(scanner.Text()))
if err := dec.Decode(&t);
err != nil {
panic(err)
}
// Perform the database operation
executionString: = "INSERT INTO observations (datetime) VALUES ($1)"
_, err := db.Exec(executionString, t.Datetime)
if err != nil {
panic(err)
}
}
My JSON file has lines, each containing a datetime value that looks like this:
{ "datetime": 1465793854 }
When the datetime is formatted as a Unix timestamp, the Marshaller complains:
panic: parsing time "1465793854" as ""2006-01-02T15:04:05Z07:00"": cannot parse "1465793854" as """
In the script that generates the JSON (also written in Golang), I tried simply printing the String representation of the Time.time type, producing the following:
{ "datetime": "2016-06-13 00:23:34 -0400 EDT" }
To which the Marshaller complains when I go to parse it:
panic: parsing time ""2016-06-13 00:23:34 -0400 EDT"" as ""2006-01-02T15:04:05Z07:00"": cannot parse " 00:23:34 -0400 EDT"" as "T"
If I also treat this timestamp (which looks pretty standard) as a String and avoid the Marshaling problem, Postgres complains when I try to perform the insertion:
panic: pq: invalid input syntax for type timestamp: "2016-06-13 00:23:34 -0400 EDT"
This is frustrating on a number of levels, but mainly because if I serialize a Time.time type, I would think it should still be understood at the other side of the process.
How can I go about parsing this timestamp to perform the database insertion? Apologizes for the lengthy question and thanks for your help!
JSON unmarshalling of time.Time expects date string to be in RFC 3339 format.
So in your golang program that generates the JSON, instead of simply printing the time.Time value, use Format to print it in RFC 3339 format.
t.Format(time.RFC3339)
if I serialize a Time.time type, I would think it should still be
understood at the other side of the process
If you used the Marshaller interface with the serializing, it would indeed output the date in RFC 3339 format. So the other side of the process will understand it. So you can do that as well.
d := DataBlob{Datetime: t}
enc := json.NewEncoder(fileWriter)
enc.Encode(d)
For reference, if you need custom unmarshalling with time types you need to create your own type with the UnmarshalJSON method. Here's an example:
type Timestamp struct {
time.Time
}
// UnmarshalJSON decodes an int64 timestamp into a time.Time object
func (p *Timestamp) UnmarshalJSON(bytes []byte) error {
// 1. Decode the bytes into an int64
var raw int64
err := json.Unmarshal(bytes, &raw)
if err != nil {
fmt.Printf("error decoding timestamp: %s\n", err)
return err
}
// 2. Parse the unix timestamp
p.Time = time.Unix(raw, 0)
return nil
}
Then use the type in your struct:
type DataBlob struct {
....
Datetime Timestamp `json:"datetime"`
....
}
Related
I am parsing rather simple json where the slash in the date format string is escaped when it arrives in response.
however, when you try to unmarshall the string, it fails on "invalid syntax" error.
So I googled and we should use strconv.Unquote to replace the escaped characters first. Did that, and now the unquote function fails on the "unknown escape" error. However JSON RFC 8259 shows it is a valid case. So why valid JSON is causing failures in Go unmarshaller?
This is the simple JSON
{
"created_at": "6\/30\/2022 21:51:49"
}
var dst string
err := json.Unmarshal([]byte("6\/30\/2022 21:51:49"), dst) // this fails on ivnalid quote
fmt.Println(dst, err)
s, err := strconv.Unquote(`"6\/30\/2022 21:51:49"`) // this fails on ivnalid syntax
fmt.Println(s, err)
How does one proceed with such cases when the JSON is valid, but none of the built-in functions of Go actually parse it? What is the right way, apart from writing simple find and replace pattern in json []byte manually?
EDIT:
ok I think I should have asked a better question or provided a specific example from the scenario rather than just a portion of it. Garbage in garbage out, my fault.
Here is the thing. Let's say we have the specific, non standard time format like in the above json example.
{
"created_at": "6\/30\/2022 21:51:49"
}
I need to parse it into the struct with go time type. Because the standard parser for time would fail for that format I create a custom type.
type Request struct {
CreatedAt Time `json:"created_at"`
}
type Time time.Time
unc (d *Time) UnmarshalJSON(b []byte) error {
s := strings.Trim(string(b), "\"")
parse, err := time.Parse("1/2/2006 15:04:05", s) // this fails because of the escaped backslashes
if err != nil {
return err
}
*d = Time(parse)
return nil
}
Here is an example in Playground: https://go.dev/play/p/bE3AWQeV-ug
panic: parsing time "6\\/30\\/2022 21:51:49" as "1/2/2006 15:04:05": cannot parse "\\/30\\/2022 21:51:49" as "/"
This is the reason I am trying to put the Unquote into the custom unmarshal function, but that is not the right approach as it was already pointed out. How does the default string json unmarshaller for example removes those escaped slashes? Oh and I have no control about the side that writes the json, it comes in with single escaped slashes like that in []byte from *http.Response
Go has a great built-in function to convert a JSON string to a Go string: json.Unmarshal. Here is how you can integrate it with a custom UnmarshalJSON method:
func (d *Time) UnmarshalJSON(b []byte) error {
var s string
if err := json.Unmarshal(b, &s); err != nil {
return err
}
// Remove this line: s := strings.Trim(string(b), "\"")
// The rest of the code is unchanged
Unquote is designed for cases when you have a double quoted string like "\"My name\"", you don't really need that in this case.
The problem with that code is that you're trying to unmarshal a byte slice that isn't JSON. You're also unmarshaling into an string which isn't going to work. You need to unmarshal into a struct that has json tags or unmarshal into a map.
var dest map[string]string
someJsonString := `{
"created_at": "6\/30\/2022 21:51:49"
}`
err := json.Unmarshal([]byte(someJsonString), &dest)
if err != nil {
panic(err)
}
fmt.Println(dest)
if you wanted to do Unmarshal into a struct, you can do it like so
type SomeStruct struct {
CreatedAt string `json:"created_at"`
}
someJsonString := `{
"created_at": "6\/30\/2022 21:51:49"
}`
var data SomeStruct
err := json.Unmarshal([]byte(someJsonString), &data)
if err != nil {
panic(err)
}
fmt.Println(data)
Currently, I try to parse JSON to map[string][]interface{}, but unmarshalling returns an error. According to (https://golang.org/pkg/encoding/json/), to unmarshal JSON into an interface value, Unmarshal stores one of these in the interface value:
bool, for JSON booleans
float64, for JSON numbers
string, for JSON strings
-[]interface{}, for JSON arrays
map[string]interface{}, for JSON objects
nil for JSON null
I wonder if golang is able to unmarshal map[string][]interface{}. The following is code snippet. I am new to Golang, thanks for help in advance.
// emailsStr looks like "{"isSchemaConforming":true,"schemaVersion":0,"unknown.0":[{"email_address":"test1#uber.com"},{"email_address":"test2#uber.com"}]}"
emailsRaw := make(map[string][]*entities.Email)
err := json.Unmarshal([]byte(emailsStr), &emailsRaw)
Error message:
&json.UnmarshalTypeError{Value:"number", Type:(*reflect.rtype)(0x151c7a0), Offset:44, Struct:"", Field:""}
The Go encoding/json package will only unmarshal dynamically to a map[string]interface{}. From there, you will need to use type assertions and casting to pull out the values you want, like so:
func main() {
jsonStr := `{"isSchemaConforming":true,"schemaVersion":0,"unknown.0":[{"email_address":"test1#uber.com"},{"email_address":"test2#uber.com"}]}`
dynamic := make(map[string]interface{})
json.Unmarshal([]byte(jsonStr), &dynamic)
firstEmail := dynamic["unknown.0"].([]interface{})[0].(map[string]interface{})["email_address"]
fmt.Println(firstEmail)
}
(https://play.golang.org/p/VEUEIwj3CIC)
Each time, Go's .(<type>) operator is used to assert and cast the dynamic value to a specific type. This particular code will panic if anything happens to be the wrong type at runtime, like if the contents of unknown.0 aren't an array of JSON objects.
The more idiomatic (and robust) way to do this in Go is to annotate a couple structs with json:"" tags and have encoding/json unmarshal into them. This avoids all the nasty brittle .([]interface{}) type casting:
type Email struct {
Email string `json:"email_address"`
}
type EmailsList struct {
IsSchemaConforming bool `json:"isSchemaConforming"`
SchemaVersion int `json:"schemaVersion"`
Emails []Email `json:"unknown.0"`
}
func main() {
jsonStr := `{"isSchemaConforming":true,"schemaVersion":0,"unknown.0":[{"email_address":"test1#uber.com"},{"email_address":"test2#uber.com"}]}`
emails := EmailsList{}
json.Unmarshal([]byte(jsonStr), &emails)
fmt.Printf("%+v\n", emails)
}
(https://play.golang.org/p/iS6e0_87P2J)
A better approach will be to use struct for main schema and then use an slice of email struct for fetching the data for email entities get the values from the same according to requirements. Please find the solution below :-
package main
import (
"fmt"
"encoding/json"
)
type Data struct{
IsSchemaConforming bool `json:"isSchemaConforming"`
SchemaVersion float64 `json:"schemaVersion"`
EmailEntity []Email `json:"unknown.0"`
}
// Email struct
type Email struct{
EmailAddress string `json:"email_address"`
}
func main() {
jsonStr := `{"isSchemaConforming":true,"schemaVersion":0,"unknown.0":[{"email_address":"test1#uber.com"},{"email_address":"test2#uber.com"}]}`
var dynamic Data
json.Unmarshal([]byte(jsonStr), &dynamic)
fmt.Printf("%#v", dynamic)
}
Background
I am learning Go and I'm trying to do some JSON unmarshaling of a datetime.
I have some JSON produced by a program I wrote in C, I am outputting what I thought was a valid ISO8601 / RFC3339 timezone offset. I'm using strftime with the following format string:
%Y-%m-%dT%H:%M:%S.%f%z
(Note that %f is not supported by strftime natively, I have a wrapper that replaces it with the nanoseconds).
This will then produce the following result:
2016-08-08T21:35:14.052975+0200
Unmarshaling this in Go however will not work:
https://play.golang.org/p/vzOXbzAwdW
package main
import (
"fmt"
"time"
)
func main() {
t, err := time.Parse(time.RFC3339Nano, "2016-08-08T21:35:14.052975+0200")
if err != nil {
panic(err)
}
fmt.Println(t)
}
Output:
panic: parsing time "2016-08-08T21:35:14.052975+0200" as "2006-01-02T15:04:05.999999999Z07:00": cannot parse "+0200" as "Z07:00"
(Working example: https://play.golang.org/p/5xcM0aHsSw)
This is because RFC3339 expects the timezone offset to be in the format 02:00 with a :, but strftime outputs it as 0200.
So I need to fix this in my C program to output the correct format.
%z The +hhmm or -hhmm numeric timezone (that is, the hour and
minute offset from UTC). (SU)
Question
However, now I have a bunch of JSON files with this incorrect format:
2016-08-08T21:35:14.052975+0200
instead of the correct (with the : in the timezone offset):
2016-08-08T21:35:14.052975+02:00
but I still want to be able to unmarshal it correctly in my Go program. Preferably two different JSON files with only this difference should parse in the exact same way.
Regarding marshaling back to JSON, the correct format should be used.
This is how I have defined it in my struct:
Time time.Time `json:"time"`
So the question is, what is the "Go" way of doing this?
Also in my code example I am using RFC3339Nano. How would I specify that in the metadata for the struct as well? As I have it now with just json:"time" will that ignore the nano seconds?
You can define your own time field type that supports both formats:
type MyTime struct {
time.Time
}
func (self *MyTime) UnmarshalJSON(b []byte) (err error) {
s := string(b)
// Get rid of the quotes "" around the value.
// A second option would be to include them
// in the date format string instead, like so below:
// time.Parse(`"`+time.RFC3339Nano+`"`, s)
s = s[1:len(s)-1]
t, err := time.Parse(time.RFC3339Nano, s)
if err != nil {
t, err = time.Parse("2006-01-02T15:04:05.999999999Z0700", s)
}
self.Time = t
return
}
type Test struct {
Time MyTime `json:"time"`
}
Try on Go Playground
In the example above we take the predefined format time.RFC3339Nano, which is defined like this:
RFC3339Nano = "2006-01-02T15:04:05.999999999Z07:00"
and remove the :
"2006-01-02T15:04:05.999999999Z0700"
This time format used by time.Parse is described here:
https://golang.org/pkg/time/#pkg-constants
Also see the documentation for time.Parse
https://golang.org/pkg/time/#Parse
P.S. The fact that the year 2006 is used in the time format strings is probably because the first version of Golang was released that year.
You can try https://play.golang.org/p/IsUpuTKENg
package main
import (
"fmt"
"time"
)
func main() {
t, err := time.Parse("2006-01-02T15:04:05.999999999Z0700", "2016-08-08T21:35:14.052975-0200")
if err != nil {
panic(err)
}
fmt.Println(t)
}
As Redis only stores strings I would like to know how I can do the equivalent of Javascript's JSON.stringify using Go to convert a Struct into a string.
I have tried typecasting:
string(the_struct)
but this results in an error.
The encoding/json package can be used to easily convert a struct to JSON string and vice versa (parse a JSON string into a struct).
Simple example (try it on the Go Playground):
type Person struct {
Name string
Age int
}
func main() {
p := Person{"Bob", 23}
// Struct -> JSON
data, err := json.Marshal(&p)
if err != nil {
panic(err)
}
fmt.Println(string(data))
// JSON -> JSON
var p2 Person
err = json.Unmarshal(data, &p2)
if err != nil {
panic(err)
}
fmt.Printf("%+v", p2)
}
Output:
{"Name":"Bob","Age":23}
{Name:Bob Age:23}
Notes:
The fields of the struct must be exported (start them with capital letter), else the json package (which uses reflection) will not be able to read/write them.
You can also specify tags for the struct fields to control/fine tune the json marshaling/unmarshaling process, for example to change the names in the JSON text:
type Person struct {
Name string `json:"name"`
Age int `json:"years"`
}
With this change the output of the above application is the following:
{"name":"Bob","years":23}
{Name:Bob Age:23}
The documentation of the json.Marshal() function details the possibilities provided by the tags.
And by implementing the json.Marshaler and json.Unmarshaler interfaces you can fully customize the marshaling / unmarshaling process.
Also if your struct is not pre-defined (e.g. you don't know the fields in advance), you can use a map[string]interface{}. See this answer for details and examples.
How can I parse this .Net JSON date with Go?
The value comes back unassigned.
It appears to parse up to the date field.
package main
import (
"encoding/json"
"fmt"
"time"
)
type MyStruct struct {
FirstField string
SomeTime time.Time
LastField string
}
type MyStructSlice struct {
MyStructs []MyStruct
}
func main() {
var s MyStructSlice
str := `{"MyStructs":[{"FirstField":"123", "SomeTime":"\/Date(1432187580000-0500)\/", "LastField":"456"}]}`
json.Unmarshal([]byte(str), &s)
fmt.Println(s)
}
Go Playground
I am going to provide a few suggestions. You will have to write the code yourself though ;)
First of all is it possible to change .NET application that produced this JSON to generate something more parsable? If you make it output datetime in RFC3339 format (something like 1990-12-31T15:59:12-08:00) then go will automatically converts it to time.Time instance thanks to http://golang.org/pkg/time/#Time.UnmarshalJSON
If you cannot change the client then you will have to parse this date yourself:
extract time part (1432187580000) from the string. this looks like number of milliseconds (ms) since UNIX epoch. You can convert it to time.Time instance using time.Unix(sec, nsec).
(optional). The time that was created in the last step already accurately represent a point in time. However if you want to add the original timezone to it (e.g. to print it) you will need to:
parse the offset part (-0500) from the string
create time.FixedZone
call http://golang.org/pkg/time/#Time.In on the instance of time.Time created in the first step
Example: http://play.golang.org/p/Pkahyg2vZa
First you model is wrong, it doesn't model the data structure in your JSON. It should be:
type Data struct {
Data []MyStruct `json:"data`
}
type MyStruct struct {
SomeTime string
}
It works with this, try it on the Go Playground.
Problem is that we still have the time as string.
Now if you want SomeTime to be time.Time, you need to parse it yourself, you can do it by implementing json.Unmarshaler:
type Data struct {
Data []MyStruct `json:"data`
}
type MyStruct struct {
SomeTime time.Time
}
func (m *MyStruct) UnmarshalJSON(data []byte) error {
// First unmashal it into a string:
ms := struct{ SomeTime string }{}
if err := json.Unmarshal(data, &ms); err != nil {
return err
}
s := ms.SomeTime
// s is of format: "/Date(1432187580000-0500)/"
// extract millis and time zone offset hours
i1 := strings.Index(s, "(")
i3 := strings.Index(s, ")")
i2 := strings.Index(s, "-")
if i2 < 0 {
i2 = strings.Index(s, "+")
}
if i1 < 0 || i2 < 0 || i3 < 0 {
return errors.New("Invalid format")
}
millis, err := strconv.ParseInt(s[i1+1:i2], 10, 64)
if err != nil {
return err
}
m.SomeTime = time.Unix(0, millis*1000000)
// Apply timezone:
zoneHours, err := strconv.ParseInt(s[i2:i3], 10, 64)
if err != nil {
return err
}
zone := time.FixedZone("something", int(zoneHours)*3600)
m.SomeTime = m.SomeTime.In(zone)
return nil
}
Try it on the Go Playground.
The value comes back unassigned for many reasons. First of all your MyStruct does not look right. Your JSON has data key as the parent key, which consists an array of objects.
But your struct for some reason does not even resembles this. It has to have Date as a field at least. So make your struct look like the json you are receiving.
Another thing is that your time.Time will not be able to parse this string: Date(1432187580000-0500) in a normal date.
P.S. now that you have updated your struct to normal way, you have to find a way to parse your strange sting to a date. If you have power over your .net application, I would rather recommend changing json to a normal timestamp or something that can be easily parsable.
If not, then you have to change SomeTime time.Time to SomeTime string and then parse string to a normal timestamp and then parse this timestamp to a date.