1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
|
<html><head><title>Advanced Redex Trails</title></head>
<body text="#000000" bgcolor="#FFFFFF" link="#0000EE" vlink="#551A8B" alink="#FF0000">
<center>
<img src="hat.gif" alt="Hat Logo"><br>
<h1>Hat</h1>
<h2>Frequently Asked Questions</h2>
</center>
<hr>
<h3>Why do you call Hat a <em>tracer</em> rather than a <em>debugger</em>?</h3>
<p>
The word <em>debugger</em> suggests: firstly that there is a fault to be
found; and secondly, that the tool will fix it. In actual fact, the
tracer is just a tool for observing what happened in any particular
program run. You can use it to find the cause of a fault, but equally,
you can use it simply to aid your understanding of a correct program.
While the tracer will help you to find bugs, it cannot find them for
you. And once the bug is found, <em>you</em> must fix it, the tracer
cannot!
<h3>Why did you not implement a conventional tracer/debugger as they
are well known for imperative languages?</h3>
<p>
Conventional tracers (debuggers) for imperative languages allow the
user to step through the program computation, stop at given points
and examine variable contents. In our opinion this tracing method is
unsuitable for a lazy functional language such as Haskell, because its
evaluation order is complex, function arguments are usually unwieldy
large unevaluated expressions, and generally computation details do
not match the user's high-level view of functions mapping values
to values. Also the stack used by the Haskell runtime system does
not resemble the stack of function calls used by runtime systems of
strict languages.
<p>
However, the trace file written by the traced program contains all
information in the order of lazy evaluation. It is possible to write
a viewing tool that admits single stepping through a lazy computation.
<h3>How much does tracing cost me in time and space?</h3>
<p>
Currently a traced program runs between 50-150 times slower than
its untraced counterpart, depending which compiler and options you
use. We are working on reducing this factor. As a rule of thumb,
a reduction step produces about 50 bytes in the trace file. So trace
files can easily grow to tens and hundreds of megabytes. However,
you can name some modules of your program as <em>trusted</em>, so
that their reductions are not recorded in the trace.
<h3>I have trouble compiling my program for tracing.</h3>
<p>
Hat covers the complete Haskell 98 language and its libraries, or
at least that is the intention. It also supports a few language
extensions, such as the FFI, multi-parameter type class, functional
dependencies, and so on. First make sure that your program is suitable
for tracing with Hat [ check our <a href="feature-table.html">Feature
Summary</a> to see exactly which language extensions are supported ],
and if it still won't compile, report it as a bug in Hat.
<p>
<hr>
The latest updates to these pages are available on the WWW from
<a href="http://www.haskell.org/hat/">
<tt>http://www.haskell.org/hat/</tt></a><br>
<a href="http://www.cs.york.ac.uk/fp/hat/">
<tt>http://www.cs.york.ac.uk/fp/hat/</tt></a>
<p>
This page last modified: 5th February 2003<br>
<a href="http://www.cs.york.ac.uk/fp/">
York Functional Programming Group</a><br>
</body></html>
|
