Moved to https://github.com/go-gorm/gorm

Make sure these boxes checked before submitting your pull request.

• [x] Do only one thing
• [x] No API-breaking changes
• [x] New code/logic commented & tested

For significant changes like big bug fixes, new features, please open an issue to make an agreement on an implementation design/plan first before starting it.

What did this pull request do?

Currently SubQuery() & QueryExpr() ignore 'gorm:query_option' which is wrong and causes sub-selects to lose locking. The patch eliminates the issue.

Make sure these boxes checked before submitting your pull request.

• [x] Do only one thing
• [x] No API-breaking changes
• [x] New code/logic commented & tested

What did this pull request do?

This will prevent creating nested transactions while calling *DB.Transaction(...) multiple times. Only fist call will create transaction. Rest will just reuse it.

Make sure these boxes checked before submitting your pull request.

• [x] Do only one thing
• [x] No API-breaking changes
• [x] New code/logic commented & tested

For significant changes like big bug fixes, new features, please open an issue to make an agreement on an implementation design/plan first before starting it.

What did this pull request do?

Make sure these boxes checked before submitting your pull request.

• [x] Do only one thing
• [x] No API-breaking changes
• [x] New code/logic commented & tested

For significant changes like big bug fixes, new features, please open an issue to make an agreement on an implementation design/plan first before starting it.

What did this pull request do?

Sometime we want to paginate convenient, we will write like below, but we will get a 0 for count:

var (
    user []User
    count int
)

if err := DB.Where("name=?", "john").Limit(10).Offset(20).Find(&users).Count(&count).Error; err != nil {
       t.Errorf("Not error should happen, but got %v", err)
}

It is because if we run Count with Offset, we will get an empty data set.

After my attempts, this can achieve the count

var (
    user []User
    count int
)

db := DB.Where("name=?", "john")
if err := db.Limit(10).Offset(20).Find(&users).Error; err != nil {
       t.Errorf("Not error should happen, but got %v", err)
}

if err := db.Count(&count).Error; err != nil {
       t.Errorf("Not error should happen, but got %v", err)
}

or

var (
    user []User
    count int
)

if err := DB.Where("name=?", "john").Limit(10).Offset(20).Find(&users).Offset(0).Count(&count).Error; err != nil {
       t.Errorf("Not error should happen, but got %v", err)
}

But the code will be more verbose. I think more people hope to complete the goal in one chain call, like here

This branch is ported from ignoreOrderQuery, use a new tag ignoreLimitAndOffsetQuery to skip the limitAndOffsetQuery

Make sure these boxes checked before submitting your pull request.

• [] Do only one thing
• [] No API-breaking changes
• [] New code/logic commented & tested

For significant changes like big bug fixes, new features, please open an issue to make an agreement on an implementation design/plan first before starting it.

What did this pull request do?

Make sure these boxes checked before submitting your pull request.

• [X] Do only one thing
• [X] No API-breaking changes
• [X] New code/logic commented & tested

For significant changes like big bug fixes, new features, please open an issue to make an agreement on an implementation design/plan first before starting it.

What did this pull request do?

Added support for setting nullable of columns using gorm.

1.Change the time to the system default time，Avoid confusion of time

Make sure these boxes checked before submitting your pull request.

• [] Do only one thing
• [] No API-breaking changes
• [] New code/logic commented & tested

For significant changes like big bug fixes, new features, please open an issue to make an agreement on an implementation design/plan first before starting it.

What did this pull request do?

• fix primaryKey DBName contains uppercase

Make sure these boxes checked before submitting your pull request.

• [X] Do only one thing
• [X] No API-breaking changes
• [X] New code/logic commented & tested

For significant changes like big bug fixes, new features, please open an issue to make an agreement on an implementation design/plan first before starting it.

What did this pull request do?

This pull request changes nothing but go.sum, removing the erroneous github.com/mattn/go-sqlite3 v2.0.1.

This PR supersedes #20. Apologies, as I chose the wrong branch to merge from.

Make sure these boxes checked before submitting your pull request.

• [x] Do only one thing
• [x] No API-breaking changes
• [x] New code/logic commented & tested

For significant changes like big bug fixes, new features, please open an issue to make an agreement on an implementation design/plan first before starting it.

What did this pull request do?

For nested preloads there seems to be no passing of errors if a preload query fails. So for example in a test case from this PR, if Level2 preload fails with error for err = localDB.Preload("Level2").Find(&got).Error, then err != nil.

However, for nested preload fails this is not the case. If Level1 preload will fail with error from the DB, for the err = localDB.Preload("Level2.Level1").Find(&got).Error query we will have err == nil. This may lead to segmentation faults when objects that are expected to be preloaded (because err == nil) are in fact not preloaded and empty.

This PR fixes that by checking currentScope errors on preload callbacks.

Make sure these boxes checked before submitting your pull request.

• [x] Do only one thing
• [x] No API-breaking changes
• [x] New code/logic commented & tested

For significant changes like big bug fixes, new features, please open an issue to make an agreement on an implementation design/plan first before starting it.

What did this pull request do?

Fixes a typo in a comment. Great package by the way!

What did this pull request do?

I'm seeing "broken pipe" errors randomly on one of our servers The issue seems to be the db driver, connections become stale when tcp connections were disconnected.

It happens more often when the DB is behind a proxy.

This has been reported and fixed in the lib/pq in v1.9+

https://github.com/lib/pq/pull/871