Differences

This shows you the differences between two versions of the page.

--- python3:input_output [2018/08/14 15:53]
kericson [Benchmarks] - minor clarification
+++ python3:input_output [2018/08/14 19:04] (current)
jguerin
@@ Line 3: / Line 3: @@
 ===== Input Basics =====
-Input in Python3 is handled via the ''%%input()%%'' function, and reads a newline-terminated string from standard input.((Unlike C++ and Java, input is line-based rather than token-based.)) Additional processing is done to the resulting string.
+Input in Python3 is handled via the ''%%input()%%'' function, and reads a newline-terminated string from standard input.((Unlike C++ and Java, input is line-based rather than token-based.)) Any additional processing is done to the resulting string.
 <code python>
@@ Line 57: / Line 57: @@
 </code>
-===== Advanced Options =====
+===== Advanced Input =====
-Like the interpreted semantics of Python3, ''%%input()%%'' and ''%%print()%%'' are slow compared to their C++ and Java counterparts.  The ''sys'' library provides faster counterparts.
+''input()'' will scale well for many easy and mid-level contest problems. For mid to upper problems with significant bounds on reads and writes ''input()'' will carry an increased risk of //time limit exceeded// judgements.((There is no guaranteed cutoff, but for many problems ''input()'' is //increasingly likely// to fail around 10<sup>3</sup>≤//n//≤10<sup>4</sup> lines.))
-<code python>
+=== stdin.readline() ===
+''readline()'' is identical to the interface provided by ''input()'' except it //retains// the ''\n'' that is used to terminate the read.((This may require special handling in certain circumstance and can be safely ignored in others. E.g., ''.split()'' will //not// behave differently vs. ''input()''.))
 <code python>
+>>> from sys import *
+>>> x = stdin.readline()
+Hello World! # x="Hello World!\n"
+</code>
-==== Benchmarks ====
-characters per line (//n//= number of lines):
-| //n//          | input() | sys.stdin.readline() | sys.stdin.readlines() |
-| 10<sup>4</sup> | .034    | .016                 | .018                  |
-| 10<sup>5</sup> | .146    | .052                 | .030                  |
-| 10<sup>6</sup> | 1.301   | .301                 | .130                  |
+=== stdin.readlines() ===
+''readlines()'' processes an entire file, terminated by an //end of file// character.((This behavior can be simulated on the terminal with ''<ctrl>+d''.))
-characters per line (//n//= number of lines):
+<code python>
-| //n//          | input() | sys.stdin.readline() | sys.stdin.readlines() |
+>>> from sys import *
-| 10<sup>4</sup> | .046    | .037                 | .033                  |
+>>> x = stdin.readlines()
-| 10<sup>5</sup> | .282    | .183                 | .143                  |
+Hello
-| 10<sup>6</sup> | 2.728   | 1.430                | 1.723                 |
+To Everyone! # x=["Hello\n", "To Everyone!\n"]
+</code>
+''<ctrl>+d'' would be used to terminate this example after the ''!''.
-The ''readlines()'' version is actually slower than ''readline()'' on the largest dataset.  It is attempting to store about 1GB in memory here, causing a slowdown, but still faster than ''input()''.
+''readlines()'' loads //and// stores an entire file into memory (as a list) its performance will exceed other options in Python3 for all but the most enormous files.((See the table below for an example where ''readlines()'' required storing ∼1GB of data.))
-<file python input.py>
-num_lines = int(input())
-for i in range(num_lines):
+==== Benchmarks ====
-    line = input()
+The following benchmarks demonstrate the increased likelihood of failure of ''input()'' as input sizes increase. All files used for testing can be found [[files:python3:io_tests|here]].
-</file>
-<file python readline.py>
+characters per line (//n//= number of lines):
-from sys import stdin
+| //n//          | input() | sys.stdin.readline() | sys.stdin.readlines() |
+| 10<sup>4</sup> | .034s    | .016s                 | .018s                  |
+| 10<sup>5</sup> | .146s    | .052s                 | .030s                  |
+| 10<sup>6</sup> | 1.301s   | .301s                 | .130s                  |
-num_lines = int(input())
-for i in range(num_lines):
+characters per line (//n//= number of lines):
-    line = stdin.readline()
+| //n//          | input() | sys.stdin.readline() | sys.stdin.readlines() |
-</file>
+| 10<sup>4</sup> | .046s    | .037s                 | .033s                  |
+| 10<sup>5</sup> | .282s    | .183s                 | .143s                  |
-<file python readlines.py>
+| 10<sup>6</sup> | 2.728s   | 1.430s                | 1.723s((The ''readlines()'' version is actually slower than ''readline()'' on the largest dataset.  It is attempting to store about 1GB in memory here, causing a slowdown, but still faster than ''input()''.
-from sys import stdin
+))                 |
-num_lines = int(input())
-lines = stdin.readlines()
-</file>
-<file python read_test.py>
+The above tests were designed to showcase minimal reading functionality other than temporary storage.((We deliberately avoided additional processing such as typecasts, ''map()'', and ''split()'', as these are non-IO considerations in Python3.))
-from sys import argv
-from random import *
-from string import *
-for i in range(int(argv[1])):
-    print(''.join(choice(ascii_lowercase + ' ') for _ in range(int(argv[2]))))
-</file>
-((''python3 read_test.py 1000 10 > out.txt''))
-((''python3 read_test.py 1000 1000 > out.txt''))

UT Martin Competitive Programming Guidebook

User Tools

Site Tools

Differences

Page Tools