1 Time-stamp: <1999-12-30 22:58:42 blp>
6 The way that data-in.c and data-out.c deal with strings is wrong. Instead of
7 the way it's done now, we should make it dynamically allocate a buffer and
8 return a pointer to it. This is a much safer interface.
10 Add libplot output driver. Suggested by Robert S. Maier
11 <rsm@math.arizona.edu>: "it produces output in idraw-editable PS format, PCL5
12 format, xfig-editable format, Illustrator format,..., and can draw vector
13 graphics on X11 displays also".
15 Storage of value labels on disk is inefficient. Invent new data structure.
17 Add an output flag which would cause a page break if a table segment could fit
18 vertically on a page but it just happens to be positioned such that it won't.
20 Fix spanned joint cells, i.e., EDLEVEL on crosstabs.stat.
24 PostScript driver should emit thin lines, then thick lines, to optimize time
27 New functions? var_name_or_label(), tab_value_or_label()
29 Should be able to bottom-justify cells. It'll be expensive, though, by
30 requiring an extra metrics call.
32 Perhaps instead of the current lines we should define the following line types:
33 null, thin, thick, double. It might look pretty classy.
35 Perhaps thick table borders that are cut off by a page break should decay to
36 thin borders. (i.e., on a thick bordered table that's longer than one page,
37 but narrow, the bottom border would be thin on the first page, and the top and
38 bottom borders on middle pages.)
40 Support multi-line titles on tables. (For the first page only, presumably.)
42 Rewrite the convert_F() function in data-out.c to be nicer code.
44 In addition to searching the source directory, we should search the current
45 directory (for data files). (Yuck!)
47 Fix line-too-long problems in PostScript code, instead of covering them up.
48 setlinecap is *not* a proper solution.
50 Need a better way than MAX_WORKSPACE to detect low-memory conditions.
52 When malloc() returns 0, page to disk and free() unnecessary data.
54 Remove ccase * argument from procfunc argument to procedure().
56 See if process_active_file() has wider applicability.
58 Looks like there's a potential problem with value labels--we use free_val_lab
59 from avl_destroy(), but free_val_lab doesn't decrement the reference count, it
60 just frees the label. Check into this sometime soon.
62 Eliminate private data in struct variable through use of pointers.
66 There needs to be another layer onto the lexer, which should probably be
67 entirely rewritten anyway. The lexer needs to read entire *commands* at a
68 time, not just a *line* at a time. This would vastly simplify the
69 (yet-to-be-implemented) logging mechanism and other stuff as well.
71 Has glob.c been pared down enough?
73 Improve interactivity of output by allowing a `commit' function for a page.
74 This will also allow for infinite-length pages.
76 All the tests need to be looked over. Some of the SET calls don't make sense
79 Implement thin single lines, should be pretty easy now.
81 SELECT IF should be moved before other transformations whenever possible. It
82 should only be impossible when one of the variables referred to in SELECT IF is
83 created or modified by a previous transformation.
85 The manual: add text, add index entries, add examples.
87 The inline file should be improved: There should be *real* detection of whether
88 it is used (in dfm.c:cmd_begin_data), not after-the-fact detection.
90 Figure out a stylesheet for messages displayed by PSPP: i.e., what quotation
91 marks around filenames, etc.
93 Data input and data output are currently arranged in reciprocal pairs: input is
94 done directly, with write_record() or whatever; output is done on a callback
95 event-driven basis. It would definitely be easier if both could be done on a
96 direct basis, with read_record() and write_record() routines, with a coroutine
97 implementation (see Knuth). But I'm not sure that coroutines can be
98 implemented in ANSI C. This will require some thought. Perhaps 0.4.0 can do
101 New SET subcommand: OUTPUT. i.e., SET OUTPUT="filename" to send output to that
102 file; SET OUTPUT="filename"(APPEND) to append to that file; SET OUTPUT=DEFAULT
103 to reset everything. There might be a better approach, though--think about it.
105 HDF export capabilities (http://hdf.ncsa.uiuc.edu). Suggested by Marcus
106 G. Daniels <mgd@santafe.edu>.
108 From Zvi Grauer <z.grauer@csuohio.edu> and <zvi@mail.ohio.net>:
110 1. design of experiments software, specifically Factorial, response surface
111 methodology and mixrture design.
113 These would be EXTREMELY USEFUL for chemists, engineeris, and anyone
114 involved in the production of chemicals or formulations.
116 2. Multidimensional Scaling analysis (for market analysis) -
118 3. Preference mapping software for market analysis
120 4. Hierarchical clustering (as well as partition clustering)
124 6. Categorical data analsys ?
129 In addition to an "infinite journal", we should keep a number of
130 individual-session journals, pspp.jnl-1 through pspp.jnl-X, renaming and
131 deleting as needed. All of the journals should have date/time comments.
133 Qualifiers for variables giving type--categorical, ordinal, ...
137 Consider consequences of xmalloc(), fail(), hcf() in interactive
139 a. Can we safely just use setjmp()/longjmp()?
140 b. Will that leak memory?
141 i. I don't think so: all procedure-created memory is either
142 garbage-collected or globally-accessible.
143 ii. But you never know... esp. w/o Checker.
144 c. Is this too early to worry? too late?
146 Need to implement a shared buffer for funny functions that require relatively
147 large permanent transient buffers (1024 bytes or so), that is, buffers that are
148 permanent in the sense that they probably shouldn't be deallocated but are only
149 used from time to time, buffers that can't be allocated on the stack because
150 they are of variable and unpredictable but usually relatively small (usually
151 line buffers). There are too many of these lurking around; can save a sizeable
152 amount of space at very little overhead and with very little effort by merging
155 Clever multiplatform GUI idea (due partly to John Williams): write a GUI in
156 Java where each statistical procedure dialog box could be downloaded from the
157 server independently. The statistical procedures would run on (the/a) server
158 and results would be reported through HTML tables viewed with the user's choice
159 of web browsers. Help could be implemented through the browser as well.
161 Design a plotting API, with scatterplots, line plots, pie charts, barcharts,
162 Pareto plots, etc., as subclasses of the plot superclass.
167 1. How to add an operator for use in PSPP expressions:
169 a. Add the operator to the enumerated type at the top of expr.h. If the
170 operator has arguments (i.e., it's not a terminal) then add it *before*
171 OP_TERMINAL; otherwise, add it *after* OP_TERMINAL. All these begin with OP_.
173 b. If the operator's a terminal then you'll want to design a structure to hold
174 its content. Add the structure to the union any_node. (You can also reuse one
175 of the prefab structures, of course.)
177 c. Now switch to expr-prs.c--the module for expression parsing. Insert the
178 operator somewhere in the precedence hierarchy.
180 (1) If you're adding a operator that is a function (like ACOS, ABS, etc.) then
181 add the function to functab in `void init_functab(void)'. Order is not
182 important here. The first element is the function name, like "ACOS". The
183 second is the operator enumerator you added in expr.h, like OP_ARCOS. The
184 third element is the C function to parse the PSPP function. The predefined
185 functions will probably suit your needs, but if not, you can write your own.
186 The fourth element is an argument to the parsing function; it's only used
187 currently by generic_str_func(), which handles a rather general syntax for
188 functions that return strings; see the comment at the beginning of its code for
191 (2) If you're adding an actual operator you'll have to put a function in
192 between two of the operators there already in functions `exprtype
193 parse_*(any_node **n)'. Each of these stores the tree for its result into *n,
194 and returns the result type, or EX_ERROR on error. Be sure to delete all the
195 allocated memory on error before returning.
197 d. Add the operator to the table `op_desc ops[OP_SENTINEL+1]' in expr-prs.c,
198 which has an entry for every operator. These entries *must* be in the same
199 order as they are in expr.h. The entries have the form `op(A,B,C,D)'. A is
200 the name of the operator as it should be printed in a postfix output format.
201 For example, the addition operator is printed as `plus'. B is a bitmapped set
204 * Set the 001 bit (OP_VAR_ARGS) if the operator takes a variable number of
205 arguments. If a function can take, say, two args or three args, but no other
206 numbers of args, this is a poor way to do it--instead implement the operator as
207 two separate operators, one with two args, the other with three. (The main
208 effect of this bit is to cause the number of arguments to be output to the
209 postfix form so that the expression evaluator can know how many args the
210 operator takes. It also causes the expression optimizer to calculate the
211 needed stack height differently, without referencing C.)
213 * Set the 002 bit (OP_MIN_ARGS) if the operator can take an optional `dotted
214 argument' that specified the minimum number of non-SYSMIS arguments in order to
215 have a non-SYSMIS result. For instance, MIN.3(e1,e2,e3,e4,e5) returns a
216 non-SYSMIS result only if at least 3 out of 5 of the expressions e1 to e5 are
219 Minargs are passed in the nonterm_node structure in `arg[]''s elements past
220 `n'--search expr-prs.c for the words `terrible crock' for an example of this.
222 Minargs are output to the postfix form. A default value is output if none was
223 specified by the user.
225 You can use minargs for anything you want--they're not limited to actually
226 describing a minimum number of valid arguments; that's just what they're most
229 * Set the 004 bit (OP_FMT_SPEC) if the operator has an argument that is a
230 format specifier. (This causes the format specifier to be output to the
231 postfix representation.)
233 Format specs are passed in the nonterm_node structure in the same way as
234 minargs, except that there are three args, in this order: type, width, # of
235 decimals--search expr-prs.c for the words `is a crock' for an example of this.
237 * Set the 010 bit (OP_ABSORB_MISS) if the operator can *ever* have a result of
238 other than SYSMIS when given one or more arguments of SYSMIS. Operators
239 lacking this bit and known to have a SYSMIS argument are short-circuited to
240 SYSMIS by the expression optimizer.
242 * If your operator doesn't fit easily into the existing categories,
243 congratulations, you get to write lots of code to adjust everything to cope
244 with this new operator. Are you really sure you want to do that?
246 C is the effect the operator has on stack height. Set this to `varies' if the
247 operator has a variable number of arguments. Otherwise this 1, minus the
248 number of arguments the operator has. (Since terminals have no arguments, they
249 have a value of +1 for this; other operators have a value of 0 or less.)
251 D is the number of items output to the postfix form after the operator proper.
252 This is 0, plus 1 if the operator has varargs, plus 1 if the operator has
253 minargs, plus 3 if the operator has a format spec. Note that minargs/varargs
254 can't coexist with a format spec on the same operator as currently coded. Some
255 terminals also have a nonzero value for this but don't fit into the above
258 e. Switch to expr-opt.c. Add code to evaluate_tree() to evaluate the
259 expression when all arguments are known to be constants. Pseudo-random
260 functions can't be evaluated even if their arguments are constants. If the
261 function can be optimized even if its arguments aren't all known constants, add
262 code to optimize_tree() to do it.
264 f. Switch to expr-evl.c. Add code to evaluate_expression() to evaluate the
265 expression. You must be absolutely certain that the code in evaluate_tree(),
266 optimize_tree(), and evaluate_expression() will always return the same results,
267 otherwise users will get inconsistent results, a Bad Thing. You must be
268 certain that even on boundary conditions users will get identical results, for
269 instance for the values 0, 1, -1, SYSMIS, or, for string functions, the null
270 string, 1-char strings, and 255-char strings.
272 g. Test the code. Write some test syntax files. Examine the output carefully.
274 NOTES ON SEARCH ALGORITHMS
275 --------------------------
277 1. Trees are nicer when you want a sorted table. However, you can always
278 sort a hash table after you're done adding values.
280 2. Brent's variation of Algorithm D is best when the table is fixed: it's
281 memory-efficient, having small, fixed overhead. It's easier to use
282 when you know in advance how many entries the table will contain.
284 3. Algorithm L is rather slow for a hash algorithm, however it's easy.
286 4. Chaining is best in terms of speed; ordered/self-ordering is even
289 5. Rehashing is slow.
291 6. Might want to decide on an algorithm empirically since there are no
292 clear mathematical winners in some cases.
294 7. gprof? Hey, it works!
296 MORE NOTES/IDEAS/BUGS
297 ---------------------
299 The behavior of converting a floating point to an integer when the value of the
300 float is out of range of the integer type is UNDEFINED! See ANSI 6.2.1.3.
302 What should we do for *negative* times in expressions?
304 Sometimes very wide (or very tall) columns can occur in tables. What is a good
305 way to truncate them? It doesn't seem to cause problems for the ascii or
306 postscript drivers, but it's not good in the general case. Should they be
307 split somehow? (One way that wide columns can occur is through user request,
308 for instance through a wide PRINT request--try time-date.stat with a narrow
309 ascii page or with the postscript driver on letter size paper.)
311 NULs in input files break the products we're replacing: although it will input
312 them properly and display them properly as AHEX format, it truncates them in A
313 format. Also, string-manipulation functions such as CONCAT truncate their
314 results after the first NUL. This should simplify the result of PSPP design.
315 Perhaps those ugly a_string, b_string, ..., can all be eliminated.
317 From Moshe Braner <mbraner@nessie.vdh.state.vt.us>: An idea regarding MATCH
318 FILES, again getting BEYOND the state of SPSS: it always bothered me that if I
319 have a large data file and I want to match it to a small lookup table, via
320 MATCH FILES FILE= /TABLE= /BY key, I need to SORT the large file on key, do the
321 match, then (usually) re-sort back into the order I really want it. There is
322 no reason to do this, when the lookup table is small. Even a dumb sequential
323 search through the table, for every case in the big file, is better, in some
324 cases, than the sort. So here's my idea: first look at the /TABLE file, if it
325 is "small enough", read it into memory, and create an index (or hash table,
326 whatever) for it. Then read the /FILE and use the index to match to each case.
327 OTOH, if the /TABLE is too large, then do it the old way, complaining if either
328 file is not sorted on key.
330 -------------------------------------------------------------------------------