1 Time-stamp: <2004-05-30 18:09:06 blp>
3 What Ben's working on now.
4 --------------------------
6 Does SET work correctly?
8 Update q2c input format description.
10 Rewrite output subsystem, break into multiple processes.
12 CROSSTABS needs to be re-examined.
14 RANK, which is needed for the Wilcoxon signed-rank statistic, Mann-Whitney U,
15 Kruskal-Wallis on NPAR TESTS and for Spearman and the Johnkheere trend test (in
21 Add compression to casefiles.
23 Expressions need to be able to abbreviate function names. XDATE.QUARTER
24 abbreviates to XDA.QUA, etc.
26 The expression tests need tests for XDATE and a few others, see
27 tests/xforms/expressions.sh comments for details.
29 Expressions need random distribution functions.
31 There needs to be another layer onto the lexer, which should probably be
32 entirely rewritten anyway. The lexer needs to read entire *commands* at a
33 time, not just a *line* at a time. It also needs to support arbitrary putback,
34 probably by just backing up the "current position" in the command buffer.
36 Scratch variables should not be available for use following TEMPORARY.
38 Details of N OF CASES, SAMPLE, FILTER, PROCESS IF, TEMPORARY, etc., need to be
39 checked against the documentation. See notes on these at end of file for a
42 Check our results against the NIST StRD benchmark results at
43 strd.itl.nist.gov/div898/strd
45 In debug mode hash table code should verify that collisions are reasonably low.
47 Use AFM files instead of Groff font files, and include AFMs for our default
48 fonts with the distribution.
50 Add libplot output driver. Suggested by Robert S. Maier
51 <rsm@math.arizona.edu>: "it produces output in idraw-editable PS format, PCL5
52 format, xfig-editable format, Illustrator format,..., and can draw vector
53 graphics on X11 displays also".
55 Storage of value labels on disk is inefficient. Invent new data structure.
57 Add an output flag which would cause a page break if a table segment could fit
58 vertically on a page but it just happens to be positioned such that it won't.
60 Fix spanned joint cells, i.e., EDLEVEL on crosstabs.stat.
64 PostScript driver should emit thin lines, then thick lines, to optimize time
67 New functions? var_name_or_label(), tab_value_or_label()
69 Should be able to bottom-justify cells. It'll be expensive, though, by
70 requiring an extra metrics call.
72 Perhaps instead of the current lines we should define the following line types:
73 null, thin, thick, double. It might look pretty classy.
75 Perhaps thick table borders that are cut off by a page break should decay to
76 thin borders. (i.e., on a thick bordered table that's longer than one page,
77 but narrow, the bottom border would be thin on the first page, and the top and
78 bottom borders on middle pages.)
80 Support multi-line titles on tables. (For the first page only, presumably.)
82 Rewrite the convert_F() function in data-out.c to be nicer code.
84 In addition to searching the source directory, we should search the current
85 directory (for data files). (Yuck!)
87 Fix line-too-long problems in PostScript code, instead of covering them up.
88 setlinecap is *not* a proper solution.
90 Need a better way than MAX_WORKSPACE to detect low-memory conditions.
92 When malloc() returns 0, page to disk and free() unnecessary data.
94 Remove ccase * argument from procfunc argument to procedure().
96 See if process_active_file() has wider applicability.
98 Eliminate private data in struct variable through use of pointers.
102 Has glob.c been pared down enough?
104 Improve interactivity of output by allowing a `commit' function for a page.
105 This will also allow for infinite-length pages.
107 All the tests need to be looked over. Some of the SET calls don't make sense
110 Implement thin single lines, should be pretty easy now.
112 SELECT IF should be moved before other transformations whenever possible. It
113 should only be impossible when one of the variables referred to in SELECT IF is
114 created or modified by a previous transformation.
116 The manual: add text, add index entries, add examples.
118 The inline file should be improved: There should be *real* detection of whether
119 it is used (in dfm.c:cmd_begin_data), not after-the-fact detection.
121 Figure out a stylesheet for messages displayed by PSPP: i.e., what quotation
122 marks around filenames, etc.
124 Data input and data output are currently arranged in reciprocal pairs: input is
125 done directly, with write_record() or whatever; output is done on a callback
126 event-driven basis. It would definitely be easier if both could be done on a
127 direct basis, with read_record() and write_record() routines, with a coroutine
128 implementation (see Knuth). But I'm not sure that coroutines can be
129 implemented in ANSI C. This will require some thought. Perhaps 0.4.0 can do
132 New SET subcommand: OUTPUT. i.e., SET OUTPUT="filename" to send output to that
133 file; SET OUTPUT="filename"(APPEND) to append to that file; SET OUTPUT=DEFAULT
134 to reset everything. There might be a better approach, though--think about it.
136 HDF export capabilities (http://hdf.ncsa.uiuc.edu). Suggested by Marcus
137 G. Daniels <mgd@santafe.edu>.
139 From Zvi Grauer <z.grauer@csuohio.edu> and <zvi@mail.ohio.net>:
141 1. design of experiments software, specifically Factorial, response surface
142 methodology and mixrture design.
144 These would be EXTREMELY USEFUL for chemists, engineeris, and anyone
145 involved in the production of chemicals or formulations.
147 2. Multidimensional Scaling analysis (for market analysis) -
149 3. Preference mapping software for market analysis
151 4. Hierarchical clustering (as well as partition clustering)
155 6. Categorical data analsys ?
160 In addition to an "infinite journal", we should keep a number of
161 individual-session journals, pspp.jnl-1 through pspp.jnl-X, renaming and
162 deleting as needed. All of the journals should have date/time comments.
164 Qualifiers for variables giving type--categorical, ordinal, ...
168 Consider consequences of xmalloc(), fail(), hcf() in interactive
170 a. Can we safely just use setjmp()/longjmp()?
171 b. Will that leak memory?
172 i. I don't think so: all procedure-created memory is either
173 garbage-collected or globally-accessible.
174 ii. But you never know... esp. w/o Checker.
175 c. Is this too early to worry? too late?
177 Need to implement a shared buffer for funny functions that require relatively
178 large permanent transient buffers (1024 bytes or so), that is, buffers that are
179 permanent in the sense that they probably shouldn't be deallocated but are only
180 used from time to time, buffers that can't be allocated on the stack because
181 they are of variable and unpredictable but usually relatively small (usually
182 line buffers). There are too many of these lurking around; can save a sizeable
183 amount of space at very little overhead and with very little effort by merging
186 Clever multiplatform GUI idea (due partly to John Williams): write a GUI in
187 Java where each statistical procedure dialog box could be downloaded from the
188 server independently. The statistical procedures would run on (the/a) server
189 and results would be reported through HTML tables viewed with the user's choice
190 of web browsers. Help could be implemented through the browser as well.
192 Design a plotting API, with scatterplots, line plots, pie charts, barcharts,
193 Pareto plots, etc., as subclasses of the plot superclass.
198 1. How to add an operator for use in PSPP expressions:
200 a. Add the operator to the enumerated type at the top of expr.h. If the
201 operator has arguments (i.e., it's not a terminal) then add it *before*
202 OP_TERMINAL; otherwise, add it *after* OP_TERMINAL. All these begin with OP_.
204 b. If the operator's a terminal then you'll want to design a structure to hold
205 its content. Add the structure to the union any_node. (You can also reuse one
206 of the prefab structures, of course.)
208 c. Now switch to expr-prs.c--the module for expression parsing. Insert the
209 operator somewhere in the precedence hierarchy.
211 (1) If you're adding a operator that is a function (like ACOS, ABS, etc.) then
212 add the function to functab in `void init_functab(void)'. Order is not
213 important here. The first element is the function name, like "ACOS". The
214 second is the operator enumerator you added in expr.h, like OP_ARCOS. The
215 third element is the C function to parse the PSPP function. The predefined
216 functions will probably suit your needs, but if not, you can write your own.
217 The fourth element is an argument to the parsing function; it's only used
218 currently by generic_str_func(), which handles a rather general syntax for
219 functions that return strings; see the comment at the beginning of its code for
222 (2) If you're adding an actual operator you'll have to put a function in
223 between two of the operators there already in functions `exprtype
224 parse_*(any_node **n)'. Each of these stores the tree for its result into *n,
225 and returns the result type, or EX_ERROR on error. Be sure to delete all the
226 allocated memory on error before returning.
228 d. Add the operator to the table `op_desc ops[OP_SENTINEL+1]' in expr-prs.c,
229 which has an entry for every operator. These entries *must* be in the same
230 order as they are in expr.h. The entries have the form `op(A,B,C,D)'. A is
231 the name of the operator as it should be printed in a postfix output format.
232 For example, the addition operator is printed as `plus'. B is a bitmapped set
235 * Set the 001 bit (OP_VAR_ARGS) if the operator takes a variable number of
236 arguments. If a function can take, say, two args or three args, but no other
237 numbers of args, this is a poor way to do it--instead implement the operator as
238 two separate operators, one with two args, the other with three. (The main
239 effect of this bit is to cause the number of arguments to be output to the
240 postfix form so that the expression evaluator can know how many args the
241 operator takes. It also causes the expression optimizer to calculate the
242 needed stack height differently, without referencing C.)
244 * Set the 002 bit (OP_MIN_ARGS) if the operator can take an optional `dotted
245 argument' that specified the minimum number of non-SYSMIS arguments in order to
246 have a non-SYSMIS result. For instance, MIN.3(e1,e2,e3,e4,e5) returns a
247 non-SYSMIS result only if at least 3 out of 5 of the expressions e1 to e5 are
250 Minargs are passed in the nonterm_node structure in `arg[]''s elements past
251 `n'--search expr-prs.c for the words `terrible crock' for an example of this.
253 Minargs are output to the postfix form. A default value is output if none was
254 specified by the user.
256 You can use minargs for anything you want--they're not limited to actually
257 describing a minimum number of valid arguments; that's just what they're most
260 * Set the 004 bit (OP_FMT_SPEC) if the operator has an argument that is a
261 format specifier. (This causes the format specifier to be output to the
262 postfix representation.)
264 Format specs are passed in the nonterm_node structure in the same way as
265 minargs, except that there are three args, in this order: type, width, # of
266 decimals--search expr-prs.c for the words `is a crock' for an example of this.
268 * Set the 010 bit (OP_ABSORB_MISS) if the operator can *ever* have a result of
269 other than SYSMIS when given one or more arguments of SYSMIS. Operators
270 lacking this bit and known to have a SYSMIS argument are short-circuited to
271 SYSMIS by the expression optimizer.
273 * If your operator doesn't fit easily into the existing categories,
274 congratulations, you get to write lots of code to adjust everything to cope
275 with this new operator. Are you really sure you want to do that?
277 C is the effect the operator has on stack height. Set this to `varies' if the
278 operator has a variable number of arguments. Otherwise this 1, minus the
279 number of arguments the operator has. (Since terminals have no arguments, they
280 have a value of +1 for this; other operators have a value of 0 or less.)
282 D is the number of items output to the postfix form after the operator proper.
283 This is 0, plus 1 if the operator has varargs, plus 1 if the operator has
284 minargs, plus 3 if the operator has a format spec. Note that minargs/varargs
285 can't coexist with a format spec on the same operator as currently coded. Some
286 terminals also have a nonzero value for this but don't fit into the above
289 e. Switch to expr-opt.c. Add code to evaluate_tree() to evaluate the
290 expression when all arguments are known to be constants. Pseudo-random
291 functions can't be evaluated even if their arguments are constants. If the
292 function can be optimized even if its arguments aren't all known constants, add
293 code to optimize_tree() to do it.
295 f. Switch to expr-evl.c. Add code to evaluate_expression() to evaluate the
296 expression. You must be absolutely certain that the code in evaluate_tree(),
297 optimize_tree(), and evaluate_expression() will always return the same results,
298 otherwise users will get inconsistent results, a Bad Thing. You must be
299 certain that even on boundary conditions users will get identical results, for
300 instance for the values 0, 1, -1, SYSMIS, or, for string functions, the null
301 string, 1-char strings, and 255-char strings.
303 g. Test the code. Write some test syntax files. Examine the output carefully.
305 NOTES ON SEARCH ALGORITHMS
306 --------------------------
308 1. Trees are nicer when you want a sorted table. However, you can always
309 sort a hash table after you're done adding values.
311 2. Brent's variation of Algorithm D is best when the table is fixed: it's
312 memory-efficient, having small, fixed overhead. It's easier to use
313 when you know in advance how many entries the table will contain.
315 3. Algorithm L is rather slow for a hash algorithm, however it's easy.
317 4. Chaining is best in terms of speed; ordered/self-ordering is even
320 5. Rehashing is slow.
322 6. Might want to decide on an algorithm empirically since there are no
323 clear mathematical winners in some cases.
325 7. gprof? Hey, it works!
327 MORE NOTES/IDEAS/BUGS
328 ---------------------
330 The behavior of converting a floating point to an integer when the value of the
331 float is out of range of the integer type is UNDEFINED! See ANSI 6.2.1.3.
333 What should we do for *negative* times in expressions?
335 Sometimes very wide (or very tall) columns can occur in tables. What is a good
336 way to truncate them? It doesn't seem to cause problems for the ascii or
337 postscript drivers, but it's not good in the general case. Should they be
338 split somehow? (One way that wide columns can occur is through user request,
339 for instance through a wide PRINT request--try time-date.stat with a narrow
340 ascii page or with the postscript driver on letter size paper.)
342 NULs in input files break the products we're replacing: although it will input
343 them properly and display them properly as AHEX format, it truncates them in A
344 format. Also, string-manipulation functions such as CONCAT truncate their
345 results after the first NUL. This should simplify the result of PSPP design.
346 Perhaps those ugly a_string, b_string, ..., can all be eliminated.
348 From Moshe Braner <mbraner@nessie.vdh.state.vt.us>: An idea regarding MATCH
349 FILES, again getting BEYOND the state of SPSS: it always bothered me that if I
350 have a large data file and I want to match it to a small lookup table, via
351 MATCH FILES FILE= /TABLE= /BY key, I need to SORT the large file on key, do the
352 match, then (usually) re-sort back into the order I really want it. There is
353 no reason to do this, when the lookup table is small. Even a dumb sequential
354 search through the table, for every case in the big file, is better, in some
355 cases, than the sort. So here's my idea: first look at the /TABLE file, if it
356 is "small enough", read it into memory, and create an index (or hash table,
357 whatever) for it. Then read the /FILE and use the index to match to each case.
358 OTOH, if the /TABLE is too large, then do it the old way, complaining if either
359 file is not sorted on key.
361 ----------------------------------------------------------------------
362 Statistical procedures:
364 For each case we read from the input program:
366 1. Execute permanent transformations. If these drop the case, stop.
367 2. N OF CASES. If we have already written N cases, stop.
368 3. Write case to replacement active file.
369 4. Execute temporary transformations. If these drop the case, stop.
370 5. Post-TEMPORARY N OF CASES. If we have already analyzed N cases, stop.
371 6. FILTER, PROCESS IF. If these drop the case, stop.
372 7. Pass case to procedure.
376 LAG records cases in step 3.
378 AGGREGATE: When output goes to an external file, this is just an ordinary
379 procedure. When output goes to the active file, step 3 should be skipped,
380 because AGGREGATE creates its own case sink and writes to it in step 7. Also,
381 TEMPORARY has no effect and we just cancel it. Regardless of direction of
382 output, we should not implement AGGREGATE through a transformation because that
383 will fail to honor FILTER, PROCESS IF, N OF CASES.
385 ADD FILES: Essentially an input program. It silently cancels unclosed LOOPs
386 and DO IFs. If the active file is used for input, then runs EXECUTE (if there
387 are any transformations) and then steals vfm_source and encapsulates it. If
388 the active file is not used for input, then it cancels all the transformations
389 and deletes the original active file.
395 MATCH FILES: Similar to AGGREGATE. This is a procedure. When the active file
396 is used for input, it reads the active file; otherwise, it just cancels all the
397 transformations and deletes the original active file. Step 3 should be
398 skipped, because MATCH FILES creates its own case sink and writes to it in step
399 7. TEMPORARY is not allowed.
407 UPDATE: same as ADD FILES.
410 ----------------------------------------------------------------------
413 * Before TEMPORARY, limits number of cases sent to the sink.
415 * After TEMPORARY, limits number of cases sent to the procedure.
417 * Without TEMPORARY, those are the same cases, so it limits both.
421 * Sample is just a transformation. It has no special properties.
425 * Always selects cases sent to the procedure.
427 * No effect on cases sent to sink.
429 * Before TEMPORARY, selection is permanent. After TEMPORARY,
430 selection stops after a procedure.
434 * Always selects cases sent to the procedure.
436 * No effect on cases sent to sink.
438 * Always stops after a procedure.
442 * Ignored by AGGREGATE. Used when procedures write matrices.
444 * Always applies to the procedure.
446 * Before TEMPORARY, splitting is permanent. After TEMPORARY,
447 splitting stops after a procedure.
451 * TEMPORARY has no effect on AGGREGATE when output goes to the active file.
453 * SORT CASES, ADD FILES, RENAME VARIABLES, CASESTOVARS, VARSTOCASES,
454 COMPUTE with a lag function cannot be used after TEMPORARY.
456 * Cannot be used in DO IF...END IF or LOOP...END LOOP.
458 * FLIP ignores TEMPORARY. All transformations become permanent.
460 * MATCH FILES and UPDATE cannot be used after TEMPORARY if active
461 file is an input source.
463 * RENAME VARIABLES is invalid after TEMPORARY.
465 * WEIGHT, SPLIT FILE, N OF CASES, FILTER, PROCESS IF apply only to
466 the next procedure when used after TEMPORARY.
470 * Always applies to the procedure.
472 * Before TEMPORARY, weighting is permanent. After TEMPORARY,
473 weighting stops after a procedure.
476 -------------------------------------------------------------------------------