
The `os` and `bufio` packages provide file reading at different granularities. `os.ReadFile` reads a whole file at once; `bufio.Scanner` reads line by line without loading the full file into memory.

<ExamplePair>
<ExampleProse>

`os.ReadFile` reads an entire file into a `[]byte` in one call. It opens, reads, and closes the file for you. Suitable for config files, small assets, and test fixtures - anything with a known bounded size.

</ExampleProse>
<ExampleCode>

```go
package main

import (
    "fmt"
    "os"
)

func main() {
    data, err := os.ReadFile("go.mod")
    if err != nil {
        panic(err)
    }
    fmt.Printf("read %d bytes\n", len(data))
    fmt.Println(string(data[:64])) // print first 64 bytes
}
```

</ExampleCode>
</ExamplePair>

<ExamplePair>
<ExampleProse>

For large files or line-oriented processing, open the file with `os.Open` and wrap it in a `bufio.Scanner`. The scanner reads one line at a time, keeping memory usage constant regardless of file size.

</ExampleProse>
<ExampleCode>

```go
package main

import (
    "bufio"
    "fmt"
    "os"
)

func main() {
    f, err := os.Open("go.mod")
    if err != nil {
        panic(err)
    }
    defer f.Close()

    scanner := bufio.NewScanner(f)
    lineNum := 0
    for scanner.Scan() {
        lineNum++
        fmt.Printf("%3d: %s\n", lineNum, scanner.Text())
    }
    if err := scanner.Err(); err != nil {
        panic(err)
    }
}
```

</ExampleCode>
</ExamplePair>

<ExamplePair>
<ExampleProse>

`os.Stdin` implements `io.Reader`, so the same `bufio.Scanner` pattern reads from standard input. This is the basis of Unix-style filter programs that accept piped input.

</ExampleProse>
<ExampleCode>

```go
package main

import (
    "bufio"
    "fmt"
    "os"
    "strings"
)

func main() {
    scanner := bufio.NewScanner(os.Stdin)
    for scanner.Scan() {
        line := scanner.Text()
        fmt.Println(strings.ToUpper(line))
    }
    if err := scanner.Err(); err != nil {
        fmt.Fprintln(os.Stderr, "read error:", err)
        os.Exit(1)
    }
}
```

</ExampleCode>
</ExamplePair>

<InProduction>
`os.ReadFile` loads the entire file into memory - fine for configs and test fixtures, unsafe for unbounded user uploads or large log files. For large or streaming reads, open with `os.Open` and wrap with `bufio.Reader` or `bufio.Scanner`. Always call `defer f.Close()` immediately after a successful `os.Open` - not after the error check, not at the end of the function, but on the very next line so it is never accidentally omitted. `bufio.Scanner` has a default line buffer of 64 KB; if you expect lines longer than that (e.g. minified JSON), use `scanner.Buffer(make([]byte, maxSize), maxSize)` before the first `Scan()`. Check `scanner.Err()` after the loop - a scanner that runs out of input and a scanner that encounters a read error both exit the loop, but only one is an error.
</InProduction>


os.ReadFile for small files, os.Open with bufio.Scanner for line-by-line reading, the io.Reader interface, and reading from stdin.

Reading Files


The `os` package provides file writing at different granularities. `os.WriteFile` handles small writes in one call; `bufio.Writer` wraps a file for efficient large or streaming output.

<ExamplePair>
<ExampleProse>

`os.WriteFile` writes a `[]byte` to a file in one call, creating or truncating the file as needed. Suitable for config files, small generated outputs, or test fixtures where the entire content is known upfront.

</ExampleProse>
<ExampleCode>

```go
package main

import (
    "fmt"
    "os"
)

func main() {
    data := []byte("hello, file\nline two\n")
    err := os.WriteFile("output.txt", data, 0644)
    if err != nil {
        panic(err)
    }
    fmt.Println("wrote", len(data), "bytes")
}
```

</ExampleCode>
</ExamplePair>

<ExamplePair>
<ExampleProse>

For large or streaming writes, create the file with `os.Create` and wrap it in a `bufio.Writer`. The writer batches small writes into larger OS calls, reducing system call overhead. Always call `Flush()` before closing.

</ExampleProse>
<ExampleCode>

```go
package main

import (
    "bufio"
    "fmt"
    "os"
)

func main() {
    f, err := os.Create("output.txt")
    if err != nil {
        panic(err)
    }
    defer f.Close()

    w := bufio.NewWriter(f)
    for i := range 5 {
        fmt.Fprintf(w, "line %d\n", i+1)
    }
    if err := w.Flush(); err != nil {
        panic(err)
    }
}
```

</ExampleCode>
</ExamplePair>

<ExamplePair>
<ExampleProse>

`os.OpenFile` gives full control over open flags. `O_APPEND` adds to an existing file instead of truncating it, `O_CREATE` creates the file if it does not exist, and `O_WRONLY` opens for writing only.

</ExampleProse>
<ExampleCode>

```go
package main

import (
    "fmt"
    "os"
)

func main() {
    f, err := os.OpenFile("log.txt",
        os.O_APPEND|os.O_CREATE|os.O_WRONLY, 0644)
    if err != nil {
        panic(err)
    }
    defer f.Close()

    _, err = fmt.Fprintln(f, "appended line")
    if err != nil {
        panic(err)
    }
}
```

</ExampleCode>
</ExamplePair>

<InProduction>
`bufio.Writer` buffers writes in memory - call `Flush()` before closing or the last buffer is silently dropped. For durable writes (config files, ledger entries), call `f.Sync()` before `f.Close()` to flush the OS page cache to disk; `defer f.Close()` alone does not guarantee durability. The safest pattern for critical writes is to write to a temp file, call `f.Sync()`, close it, then rename atomically with `os.Rename` - readers either see the old file or the complete new file, never a partial write. `os.OpenFile` with `O_APPEND` is not atomic for writes larger than the OS write size on some filesystems; for append-only logs under concurrent writers, use a dedicated logging library or serialize writes through a goroutine.
</InProduction>


os.WriteFile for small writes, os.Create with bufio.Writer for large or streaming writes, os.OpenFile with flags, and fsync for durability.

Writing Files


The `encoding/base64` package encodes binary data as printable ASCII text. Two encodings are available: `StdEncoding` (uses `+` and `/`) and `URLEncoding` (uses `-` and `_`, safe in URLs).

<ExamplePair>
<ExampleProse>

`base64.StdEncoding.EncodeToString` encodes a byte slice to a base64 string. `DecodeString` reverses it, returning a byte slice and an error if the input is malformed.

</ExampleProse>
<ExampleCode>

```go
package main

import (
    "encoding/base64"
    "fmt"
)

func main() {
    data := []byte("user:secret-password")

    encoded := base64.StdEncoding.EncodeToString(data)
    fmt.Println(encoded) // dXNlcjpzZWNyZXQtcGFzc3dvcmQ=

    decoded, err := base64.StdEncoding.DecodeString(encoded)
    if err != nil {
        panic(err)
    }
    fmt.Println(string(decoded)) // user:secret-password
}
```

</ExampleCode>
</ExamplePair>

<ExamplePair>
<ExampleProse>

`base64.URLEncoding` replaces `+` with `-` and `/` with `_`, making the output safe to embed in URLs and filenames. `base64.RawURLEncoding` omits the `=` padding - used in JWT segments and URL-safe tokens.

</ExampleProse>
<ExampleCode>

```go
package main

import (
    "crypto/rand"
    "encoding/base64"
    "fmt"
)

func main() {
    // Standard encoding - may produce + and / in output
    std := base64.StdEncoding.EncodeToString([]byte{0xfb, 0xef, 0xff})
    fmt.Println("std:", std) // ++//

    // URL-safe encoding - replaces + with - and / with _
    url := base64.URLEncoding.EncodeToString([]byte{0xfb, 0xef, 0xff})
    fmt.Println("url:", url) // --__

    // Generate a URL-safe random token (no padding)
    buf := make([]byte, 24)
    rand.Read(buf)
    token := base64.RawURLEncoding.EncodeToString(buf)
    fmt.Println("token:", token) // 32-char URL-safe token, no = padding
}
```

</ExampleCode>
</ExamplePair>

<ExamplePair>
<ExampleProse>

For large payloads, wrap an `io.Writer` with `base64.NewEncoder` to stream encoding without buffering the entire output. Call `Close()` to flush any remaining bytes and write the final padding.

</ExampleProse>
<ExampleCode>

```go
package main

import (
    "encoding/base64"
    "os"
    "strings"
)

func main() {
    enc := base64.NewEncoder(base64.StdEncoding, os.Stdout)
    defer enc.Close()

    r := strings.NewReader("streaming base64 example output")
    buf := make([]byte, 8)
    for {
        n, err := r.Read(buf)
        if n > 0 {
            enc.Write(buf[:n])
        }
        if err != nil {
            break
        }
    }
}
```

</ExampleCode>
</ExamplePair>

<InProduction>
Use `base64.URLEncoding` (or `base64.RawURLEncoding` without padding) for URL-safe tokens and JWT segments - standard base64 uses `+` and `/` which are special characters in URLs and require percent-encoding. When size matters, note that base64 inflates data by approximately 33% (3 bytes become 4 characters). For generating random tokens (session IDs, CSRF tokens, API keys), fill a byte slice with `crypto/rand.Read`, then `base64.RawURLEncoding.EncodeToString` - 18 random bytes give 24 URL-safe characters with 144 bits of entropy, sufficient for most token use cases. Never roll a custom encoding for tokens; use the stdlib encoders to avoid subtle bugs with alphabet selection and padding.
</InProduction>


encoding/base64 StdEncoding and URLEncoding for encode/decode, RawURLEncoding for padding-free tokens, and streaming encoders.