Information Technology Reference
In-Depth Information
T
ABLE
11.
Resources used during compressionandindexing.Only the vocab-
ulary is constructed in thefirstpass; theother structures are built in the second
pass.
Pass
Output
Size
CPUM m
Mb
%
Hr:Min
Mb
Pass 1:
Compression
Model
4.2
0.2
2:37
25.6
Inversion
Vocabulary
6.4
0.3
3:02
18.7
Overhead
0:19
2.5
Total
10.6
0.5
5:58
46.8
Pass 2:
Compression
Text
605.1
29.4
3:27
25.6
Doc. map
2.8
0.1
Inversion
Index
132.2
6.4
5:25
162.1
Index map
2.1
0.1
Doc. lens
2.8
0.1
Approx. lens
0.7
0.0
Overhead
0:23
2.5
Total
745.8
36.3
9:15
190.2
Overall
756.4
36.8
15:13
190.2
Fig. 11.15
Jumbled table. Columns have been crammed together and are hard to understand. The
numbers representing quantities of the same kind don't line up vertically. The percentage column is
mysterious, since it doesn't sum to 100. It seems unlikely that all the detail is interesting; consider
in particular the “Index map”, “Doc. lens”, and “Appr. lens” rows, which could presumably be
gathered into a single row with a label such as “Other” or discarded altogether
As illustrated in Table 6, temporary space requirements were 60 % to 65 % of
the data size.
In our experiments, temporary space requirements were 60 % to 65 % of the
data size.
Small tables can be part of the running text, displayed in the same way as mathematics.
Larger tables should be labelled and positioned at the top or bottom of a page.
Tables are used not only for numbers but for analysis of alternatives. For example,
a list of approaches to system modelling could be compared in a table, one row
per approach, with columns used for positives, negatives, and number of known
successful applications of the approach. In such tables, each cell may contain a brief
paragraph of text and a single table may occupy a page or more, and thus the overall
appearance is quite different to that of a table of numerical data. Nonetheless, the
same design guidelines apply.
