Golang and MySQL login tutorial | Hacker News Reader

Golang and MySQL login tutorial

dinosaurscode.xyz

56 pointslllorddino17 comments

    hashPassword, _ := bcrypt.GenerateFromPassword([]byte(password), bcrypt.DefaultCost)

    _, err = db.Exec("INSERT INTO users(usrname, password) VALUES(?, ?)", username, password)

I'm not a Go expert. Doesn't this:

1) ignore an error, if any? And then

2) ignore hashPassword and just insert the non-hashed password into the database?

It looks like if someone follows along from home they'll get a syntax error from "usrname" but let's say they fixed that...

The code at the bottom of the article doesn't seem to suffer from the latter issue, but still suffers from the former.

Seems like the author went a little too fast here. Should definitely insert the hashed password into the db.

edit: Otherwise, great read! Hoping for more.

Yeah, it seems like a typo. The "Signup Page" example doesn't use `hashPassword` but in the "entire source code" example at the bottom, it does properly use `hashedPassword` in the statement. (But still ignores any error from `bcrypt.GenerateFromPassword`)

Fixed and fixed. There were more bugs in it but glad you pointed this out or else I may have never fixed it.

panic() in a web serving route (e.g. signupPage here) is bad practice - an error will bring down your web server. Panics should be reserved for errors that invalidate your system, e.g. missing view assets or db connection on system startup.

Much better to http.Redirect/5xx.

Fixed.

I'm not familiar with how the DB classes work (and I guess what counts as a function "ending"). Since `db.Close()` is deferred, is there a mysql connection open for as long as the web server is open in this case?

The main() function doesn't "end" until the web listener is stopped, correct?

Yes, that's correct. `http.ListenAndServe()` doesn't return until the web listener is stopped.

This is actually the intended usage of `sql.Open()`: https://golang.org/pkg/database/sql/#Open (it maintains its own connection pool and is safe for concurrent use)

Whilst I am always grateful for new tutorials, I think this is a special case where the author should have gone the extra mile and talk about salting and maybe even about hashing the password client side.

Security related tutorials should not skip steps, even if only to mention the more advanced of them.

There is no skipped step, bcrypt does salting itself.

Ah yes! Thank you for the correction. :)

    var db *sql.DB
    var err error

Why declare this error globally?

The DB: because there's an internal goroutine-safe connection pool, so you want it to either be global, or exist in the server singleton (i.e. global).

The err: because they use it so frequently, easier to define it once and simply assign to it multiple times than try and track if it's already been defined per block of code.

Personally, I prefer to use distinct error variables for each error which can be thrown, but either method is idiomatic.

It's bad practice to define an error globally. Let's say you create a goroutine which uses that same err variable (such as an http handler)--now you have a race condition, as that memory is being modified across threads.

True, and looking at the code, it doesn't make sense to define it as a global, since each potential goroutine (i.e. the HTTP handler functions) is re-defining 'err' whenever it's being used.

why not use an ORM?

It's a good practice to learn the inner works before adding layers that may simplify (but obscure) functionality.

Reply on news.ycombinator.com