Readers, records, and partial progress

The idea

The io.Reader interface reports both a byte count and an error. A call may return useful bytes together with a terminal error, so process the first n bytes before handling that error. Only the returned prefix belongs to the current read. Buffer capacity does not tell you how much fresh data arrived, and a later read may overwrite the same storage.

bufio.Scanner offers convenient token iteration, including line scanning by default. After its loop ends, check Err to distinguish ordinary completion from a scanning failure. Scanner also has a token-size limit unless configured differently. At the record layer, decide whether malformed input stops the operation or is collected as a separate failure; silently treating bad data as zero loses important information.

package main
import ("fmt"; "bufio"; "strings")
func main(){ s:=bufio.NewScanner(strings.NewReader("red

blue
")); n:=0; for s.Scan(){ if s.Text()!="" {n++} }; if err:=s.Err(); err!=nil {panic(err)}; fmt.Println(n) }

2

Three lines are scanned; the explicit empty-line policy counts only red and blue, and scanner failure is checked afterward.