Benchmark: timing 100 iterations of html, join, local, manual... html: 7 wallclock secs ( 6.36 usr + 0.01 sys = 6.37 CPU) @ 15.70/s (n=100) join: 22 wallclock secs (22.27 usr + 0.01 sys = 22.28 CPU) @ 4.49/s (n=100) local: 22 wallclock secs (22.46 usr + 0.02 sys = 22.48 CPU) @ 4.45/s (n=100) manual: 76 wallclock secs (75.38 usr + 0.01 sys = 75.39 CPU) @ 1.33/s (n=100)