I had promised that the *.dat files would never change, but I learned later that miles.dat as originally distributed used the wrong two-letter state codes for Nebraska and Saskatchewan. This was politically incorrect. I decided to change to the correct codes, since this won't affect any of the data that's critical for the graph structure. Please replace your copies of miles.dat (and cities.texmap) if they are dated earlier than 1995. Similarly, I found an embarrassing typo ("Monsiuer" for "Monsieur") in the file jean.dat; and there was a spurious white pixel at the very bottom right corner of lisa.dat; those glitches were corrected in March 2007. ---------------------------------------- Here is a list of all significant changes that appeared in the third printing of The Stanford GraphBase book (2002). Corrections marked with * were already made in the second printing (1994); see the webpage http://www-cs-faculty.stanford.edu/~knuth/sgb.html for corrections to any subsequent printings. If you know of an error not listed here or there, chances are good that you can earn a reward by reporting it to sgb@cs.stanford.edu. [Note that comparison between pointers from differently calloc'd regions of memory is not considered erroneous; see the README file.] Many thanks to the readers who reported these glitches. Page 1, line 17: should say The eleven data files *.dat occupy 355,156 bytes Page 7, line -11: change "rough, black, empty" to "chump, given, flaky" Page 7, line -8: change "alphabet." to "alphabet. One solution, namely {murky, qophs, vixen, waltz}, hit 14 of the last 16." *Page 10, line 9: should say John Lewis Roget *Page 14, line 6: change "it is" to "is" Page 20, line 1: change "&" to "\&" Page 21, lines 16, 15, 2, and -1: change 11519, 17090, 14650, 23098 to 11736, 17770, 14919, 23901, respectively *Pages 22 and 24: the illustrations are too low on the page Page 22: should say Regina, SK and Scottsbluff, NE Page 23, lines 6 and 7: change 91256 to 91953 and 274127 to 290169 Page 26, insert at beginning of line 17: (In rare cases, the triangle inequality will fail.) *Page 26, line 17: $2^14$ is, of course, 16384 Page 28, line 2: should say "is much more difficult than the" *Pages 29 and 30: change "dominos" to "dominoes" (7 times) Page 33, line 14: insert thin space before "..." *Page 34: remove extra comma in the caption to Figure 11 Page 34, caption to Figure 12: change "(28,4,8,4,0,0)" to "(8,4,8,4,0,0,0)" *Page 40: remove the first "poset," on line 19 Page 43, line 5: change "1982 and 1984" to "1984 and 1986" Page 43, line 7: change "1988" to "1989" Page 45, line 12: change "under under" to "under" Page 45, line 23: change "279" to "239" Page 46, line 4 from the bottom: industry $j$, not industry $k$ Page 52, line 5: change "the vertices" to "two vertices" *Page 58: numbered line 61 should end "+The+Stanford+GraphBase+" *Page 58: numbered lines 69 and 70 should end with double quotes *Page 59: insert change "test" to "./test" on numbered lines 87, 88, 89, 93 Page 60: change "-o foo" to "-o foo foo.c" in both displayed command lines Page 60, line 7 from bottom: change "/usr" to "`/usr" Page 61, line 5 from bottom: change "CHANGEFILE" to "CHANGEFILES" Page 62: line 25: change "lines 92--96" to "lines 73--77" Page 63, line 15: change "gb_graph.o" to "gb_graph.o," Page 75: remove semicolon at end of line 2 Page 78, line 8: The accent in K\"onig should actually be the long Hungarian accent, K\H{o}nig. (Also on line 6 from the bottom, and on page 79 lines 1 and 11, and in the index on page 573.) Page 78: the long-double-dash notation used on lines 15 and following is changed to a heavy bar Page 78: in display (*) and the following lines, use notation like $r^{(0)$ and $c^{(1)}$ instead of $r_0$ and $c_1$ Page 78, line 7 from bottom: change "matrix" to "matrix, of size $n\times n$" Page 81, line 14: change "row $l$" to "column $l$" *Page 90, lines -10 and -9: remove the semicolons *Page 99, line 6: change "x --- y" to "y --- z" *Page 121, line 4 should say: "assume that |BUF_SIZE| is rather large, but in cases of doubt they ensure that |BUF_SIZE| will never be exceeded." Page 124, line 11 from bottom: change "x_k+\delta_k" to "x_d+\delta_d" Page 138, line 15 should reverse the order of the tests: if (ss: also say #include (twice) *Page 287, insert a new paragraph after line 4, to wit: {\sl Important Note:\/} Programs of the Stanford GraphBase implicitly assume that all memory allocated by |calloc| comes from a single underlying memory array. Therefore pointer values are compared to each other in many places, even when the objects pointed to have been allocated at different times. Strictly speaking, this liberal use of pointer comparisons fails to conform to the restrictions of ANSI Standard \CEE/, when the comparison involves a less-than or greater-then relation. Users whose system supports only the strict standard will need to make several dozen changes. *Page 288, section 18, line 11: replace "g->uu.I++;" by the code { g->uu.I++; printf("."); fflush(stdout); } and make entries for fflush and stdout in the mini-index on page 289 Page 303, line 2: change "check sum" to "checksum" *Page 304, line 8: change "is none" to "if none" Page 304, line 20: change "81" to "sizeof(buffer)" Page 305, line 15: change 100 to |unexpected_char| Page 306, bottom line: change "icode[*p]" to "imap_ord(*p)" Page 308, replacement for 3rd line from bottom: icode[0]=d; /* make sure '\0' is a nondigit */ if (imap_ord(*cur_pos)= static char item_buf[MAX_SV_STRING+3+81]; /* an item to be output */ Page 413, line 10: change "output; in" to "output; the value $-3$ will be returned if memory is exhausted. In" Page 414, line 5: remove the semicolon Page 414, swap lines 7--8 with with lines 9--10 *Page 414: last line should end ", or if comparisons between pointers are undefined." Page 415: new line of code to follow line 12: if (blocks==NULL) return -3; /* out of memory */ Page 426, line 13: change "common words" to "common five-letter words" Page 430: the bottom few lines should move to page 431 Page 436, section 17: node_blocks should be declared static (this affects the mini-indexes on pages 433 and 439) Page 442, line 23: change "3,675,808" to "3,674,808" *Page 444, section 6, line 4: insert "n-vertex" before "regular" *Page 444, section 7, line 3: should be superscript g, not subscript Page 445, line -6: change "4*q*2" to "4*q*q" Pages 452 and 453: change "rand" to "randm" (six times) Page 456, lines 8 and 9 of section 18: change "u->arcs->len" to "(u->arcs-1)->len" and vice versa *Page 461, line 16: remove the semicolon Page 466, line 12 from the bottom: remove the semicolon Page 481, line 12: change "u->parent" to "o,u->parent" Page 490, line 19: remove the semicolon Page 494, line 5 of section 65: change "t->rsib" to "o,t->rsib" Page 495, line 5: change "o,k" to "oo,k" *Page 497, line 2: change "respect" to "respect to" Page 497, line 18: change "1000" to "100" Page 497, lines 16, 17, 29, 30: change "11519" to "11736", "17090" to "17770", "58544" to "59050", and "165749" to "175519" *Page 512, section 2, line 9: remove the semicolon *Page 537: add "GB_GRAPH 18," to the index listings for fflush Page 539: icode is not in GB_IO 17 Page 541: missing_operand is in GB_BASIC 81,87, 95 Page 542: no_room is not in GB_BASIC 81,87, 95 *Page 543: section numbers in the index listings for panic_code should not be underlined except in GB_GRAPH Page 544: rand becomes randm (and alphabetized later) *Page 545: add "GB_GRAPH 18," to the index listings for stdout Page 547: add section 10 to the index listings for unexpected_char in GB_IO Page 555: move the prototype for gb_number to page 554, where it belongs in alphabetic order Page 557: "all_parts(10)" should be "all_parts(10,0)" Page 558: "all_perms(5)" should be "all_perms(5,0)" Page 559: "book("homer",280,0,1,0,0,1,1,0)" should be "book("homer",280,0,0,0,1,1,0)" Page 561: "all_perms(7)" should be "all_perms(7,0)" Page 561, last two lines: change "multiple edges" to "multiple arcs" Page 562, lines 7 and 11: change "multiple edges" to "multiple arcs" Page 568, line 5: change ". 15" to ", 15" Pages 568 and 569: insert thin space after the comma when giving issue numbers, in references [5], [9], [15], [16], [23], [30], [31]. *Page 572: "Jarn{\'\i}, k" should be "Jarn{\'\i}k" Page 573: "K\"onig" should be "K\H{o}nig" Page 573: should say "literate programming" not "literate programmming" Page 573, Marsaglia, George, 217. Page 574: permutations, random, 104. *Page 574: should say Roget, John Lewis Page 575: should say "Tolsto\u\i, Lev Nikolaevich" Page 575: add page 26 under triangle inequality Page 575: Voronoi's correct name is "Vorono\u\i, Georgi\u\i\ Feodos'evich" Page 575: Walker's full name is "Walker, Alistair John" Refinements to spacing between lines, relevant to section names like , have also been made on pages 75, 124, 131, 137, 140, 146, 150, 156, 158, 160, 163, 168, 174, 182, 204, 225, 318, 324, 329, 334, 367, 399, 414, 433, 466, 490.