1 Time-stamp: <2004-03-20 19:16:04 blp>
6 There needs to be another layer onto the lexer, which should probably be
7 entirely rewritten anyway. The lexer needs to read entire *commands* at a
8 time, not just a *line* at a time. It also needs to support arbitrary putback,
9 probably by just backing up the "current position" in the command buffer.
11 Scratch variables should not be available for use following TEMPORARY.
13 Details of N OF CASES, SAMPLE, FILTER, PROCESS IF, TEMPORARY, etc., need to be
14 checked against the documentation. See notes on these at end of file for a
17 Check our results against the NIST StRD benchmark results at
18 strd.itl.nist.gov/div898/strd
20 In debug mode hash table code should verify that collisions are reasonably low.
22 Use posix_fadvise(POSIX_FADV_SEQUENTIAL) where available.
24 Use AFM files instead of Groff font files, and include AFMs for our default
25 fonts with the distribution.
27 Add libplot output driver. Suggested by Robert S. Maier
28 <rsm@math.arizona.edu>: "it produces output in idraw-editable PS format, PCL5
29 format, xfig-editable format, Illustrator format,..., and can draw vector
30 graphics on X11 displays also".
32 Storage of value labels on disk is inefficient. Invent new data structure.
34 Add an output flag which would cause a page break if a table segment could fit
35 vertically on a page but it just happens to be positioned such that it won't.
37 Fix spanned joint cells, i.e., EDLEVEL on crosstabs.stat.
41 PostScript driver should emit thin lines, then thick lines, to optimize time
44 New functions? var_name_or_label(), tab_value_or_label()
46 Should be able to bottom-justify cells. It'll be expensive, though, by
47 requiring an extra metrics call.
49 Perhaps instead of the current lines we should define the following line types:
50 null, thin, thick, double. It might look pretty classy.
52 Perhaps thick table borders that are cut off by a page break should decay to
53 thin borders. (i.e., on a thick bordered table that's longer than one page,
54 but narrow, the bottom border would be thin on the first page, and the top and
55 bottom borders on middle pages.)
57 Support multi-line titles on tables. (For the first page only, presumably.)
59 Rewrite the convert_F() function in data-out.c to be nicer code.
61 In addition to searching the source directory, we should search the current
62 directory (for data files). (Yuck!)
64 Fix line-too-long problems in PostScript code, instead of covering them up.
65 setlinecap is *not* a proper solution.
67 Need a better way than MAX_WORKSPACE to detect low-memory conditions.
69 When malloc() returns 0, page to disk and free() unnecessary data.
71 Remove ccase * argument from procfunc argument to procedure().
73 See if process_active_file() has wider applicability.
75 Eliminate private data in struct variable through use of pointers.
79 Has glob.c been pared down enough?
81 Improve interactivity of output by allowing a `commit' function for a page.
82 This will also allow for infinite-length pages.
84 All the tests need to be looked over. Some of the SET calls don't make sense
87 Implement thin single lines, should be pretty easy now.
89 SELECT IF should be moved before other transformations whenever possible. It
90 should only be impossible when one of the variables referred to in SELECT IF is
91 created or modified by a previous transformation.
93 The manual: add text, add index entries, add examples.
95 The inline file should be improved: There should be *real* detection of whether
96 it is used (in dfm.c:cmd_begin_data), not after-the-fact detection.
98 Figure out a stylesheet for messages displayed by PSPP: i.e., what quotation
99 marks around filenames, etc.
101 Data input and data output are currently arranged in reciprocal pairs: input is
102 done directly, with write_record() or whatever; output is done on a callback
103 event-driven basis. It would definitely be easier if both could be done on a
104 direct basis, with read_record() and write_record() routines, with a coroutine
105 implementation (see Knuth). But I'm not sure that coroutines can be
106 implemented in ANSI C. This will require some thought. Perhaps 0.4.0 can do
109 New SET subcommand: OUTPUT. i.e., SET OUTPUT="filename" to send output to that
110 file; SET OUTPUT="filename"(APPEND) to append to that file; SET OUTPUT=DEFAULT
111 to reset everything. There might be a better approach, though--think about it.
113 HDF export capabilities (http://hdf.ncsa.uiuc.edu). Suggested by Marcus
114 G. Daniels <mgd@santafe.edu>.
116 From Zvi Grauer <z.grauer@csuohio.edu> and <zvi@mail.ohio.net>:
118 1. design of experiments software, specifically Factorial, response surface
119 methodology and mixrture design.
121 These would be EXTREMELY USEFUL for chemists, engineeris, and anyone
122 involved in the production of chemicals or formulations.
124 2. Multidimensional Scaling analysis (for market analysis) -
126 3. Preference mapping software for market analysis
128 4. Hierarchical clustering (as well as partition clustering)
132 6. Categorical data analsys ?
137 In addition to an "infinite journal", we should keep a number of
138 individual-session journals, pspp.jnl-1 through pspp.jnl-X, renaming and
139 deleting as needed. All of the journals should have date/time comments.
141 Qualifiers for variables giving type--categorical, ordinal, ...
145 Consider consequences of xmalloc(), fail(), hcf() in interactive
147 a. Can we safely just use setjmp()/longjmp()?
148 b. Will that leak memory?
149 i. I don't think so: all procedure-created memory is either
150 garbage-collected or globally-accessible.
151 ii. But you never know... esp. w/o Checker.
152 c. Is this too early to worry? too late?
154 Need to implement a shared buffer for funny functions that require relatively
155 large permanent transient buffers (1024 bytes or so), that is, buffers that are
156 permanent in the sense that they probably shouldn't be deallocated but are only
157 used from time to time, buffers that can't be allocated on the stack because
158 they are of variable and unpredictable but usually relatively small (usually
159 line buffers). There are too many of these lurking around; can save a sizeable
160 amount of space at very little overhead and with very little effort by merging
163 Clever multiplatform GUI idea (due partly to John Williams): write a GUI in
164 Java where each statistical procedure dialog box could be downloaded from the
165 server independently. The statistical procedures would run on (the/a) server
166 and results would be reported through HTML tables viewed with the user's choice
167 of web browsers. Help could be implemented through the browser as well.
169 Design a plotting API, with scatterplots, line plots, pie charts, barcharts,
170 Pareto plots, etc., as subclasses of the plot superclass.
175 1. How to add an operator for use in PSPP expressions:
177 a. Add the operator to the enumerated type at the top of expr.h. If the
178 operator has arguments (i.e., it's not a terminal) then add it *before*
179 OP_TERMINAL; otherwise, add it *after* OP_TERMINAL. All these begin with OP_.
181 b. If the operator's a terminal then you'll want to design a structure to hold
182 its content. Add the structure to the union any_node. (You can also reuse one
183 of the prefab structures, of course.)
185 c. Now switch to expr-prs.c--the module for expression parsing. Insert the
186 operator somewhere in the precedence hierarchy.
188 (1) If you're adding a operator that is a function (like ACOS, ABS, etc.) then
189 add the function to functab in `void init_functab(void)'. Order is not
190 important here. The first element is the function name, like "ACOS". The
191 second is the operator enumerator you added in expr.h, like OP_ARCOS. The
192 third element is the C function to parse the PSPP function. The predefined
193 functions will probably suit your needs, but if not, you can write your own.
194 The fourth element is an argument to the parsing function; it's only used
195 currently by generic_str_func(), which handles a rather general syntax for
196 functions that return strings; see the comment at the beginning of its code for
199 (2) If you're adding an actual operator you'll have to put a function in
200 between two of the operators there already in functions `exprtype
201 parse_*(any_node **n)'. Each of these stores the tree for its result into *n,
202 and returns the result type, or EX_ERROR on error. Be sure to delete all the
203 allocated memory on error before returning.
205 d. Add the operator to the table `op_desc ops[OP_SENTINEL+1]' in expr-prs.c,
206 which has an entry for every operator. These entries *must* be in the same
207 order as they are in expr.h. The entries have the form `op(A,B,C,D)'. A is
208 the name of the operator as it should be printed in a postfix output format.
209 For example, the addition operator is printed as `plus'. B is a bitmapped set
212 * Set the 001 bit (OP_VAR_ARGS) if the operator takes a variable number of
213 arguments. If a function can take, say, two args or three args, but no other
214 numbers of args, this is a poor way to do it--instead implement the operator as
215 two separate operators, one with two args, the other with three. (The main
216 effect of this bit is to cause the number of arguments to be output to the
217 postfix form so that the expression evaluator can know how many args the
218 operator takes. It also causes the expression optimizer to calculate the
219 needed stack height differently, without referencing C.)
221 * Set the 002 bit (OP_MIN_ARGS) if the operator can take an optional `dotted
222 argument' that specified the minimum number of non-SYSMIS arguments in order to
223 have a non-SYSMIS result. For instance, MIN.3(e1,e2,e3,e4,e5) returns a
224 non-SYSMIS result only if at least 3 out of 5 of the expressions e1 to e5 are
227 Minargs are passed in the nonterm_node structure in `arg[]''s elements past
228 `n'--search expr-prs.c for the words `terrible crock' for an example of this.
230 Minargs are output to the postfix form. A default value is output if none was
231 specified by the user.
233 You can use minargs for anything you want--they're not limited to actually
234 describing a minimum number of valid arguments; that's just what they're most
237 * Set the 004 bit (OP_FMT_SPEC) if the operator has an argument that is a
238 format specifier. (This causes the format specifier to be output to the
239 postfix representation.)
241 Format specs are passed in the nonterm_node structure in the same way as
242 minargs, except that there are three args, in this order: type, width, # of
243 decimals--search expr-prs.c for the words `is a crock' for an example of this.
245 * Set the 010 bit (OP_ABSORB_MISS) if the operator can *ever* have a result of
246 other than SYSMIS when given one or more arguments of SYSMIS. Operators
247 lacking this bit and known to have a SYSMIS argument are short-circuited to
248 SYSMIS by the expression optimizer.
250 * If your operator doesn't fit easily into the existing categories,
251 congratulations, you get to write lots of code to adjust everything to cope
252 with this new operator. Are you really sure you want to do that?
254 C is the effect the operator has on stack height. Set this to `varies' if the
255 operator has a variable number of arguments. Otherwise this 1, minus the
256 number of arguments the operator has. (Since terminals have no arguments, they
257 have a value of +1 for this; other operators have a value of 0 or less.)
259 D is the number of items output to the postfix form after the operator proper.
260 This is 0, plus 1 if the operator has varargs, plus 1 if the operator has
261 minargs, plus 3 if the operator has a format spec. Note that minargs/varargs
262 can't coexist with a format spec on the same operator as currently coded. Some
263 terminals also have a nonzero value for this but don't fit into the above
266 e. Switch to expr-opt.c. Add code to evaluate_tree() to evaluate the
267 expression when all arguments are known to be constants. Pseudo-random
268 functions can't be evaluated even if their arguments are constants. If the
269 function can be optimized even if its arguments aren't all known constants, add
270 code to optimize_tree() to do it.
272 f. Switch to expr-evl.c. Add code to evaluate_expression() to evaluate the
273 expression. You must be absolutely certain that the code in evaluate_tree(),
274 optimize_tree(), and evaluate_expression() will always return the same results,
275 otherwise users will get inconsistent results, a Bad Thing. You must be
276 certain that even on boundary conditions users will get identical results, for
277 instance for the values 0, 1, -1, SYSMIS, or, for string functions, the null
278 string, 1-char strings, and 255-char strings.
280 g. Test the code. Write some test syntax files. Examine the output carefully.
282 NOTES ON SEARCH ALGORITHMS
283 --------------------------
285 1. Trees are nicer when you want a sorted table. However, you can always
286 sort a hash table after you're done adding values.
288 2. Brent's variation of Algorithm D is best when the table is fixed: it's
289 memory-efficient, having small, fixed overhead. It's easier to use
290 when you know in advance how many entries the table will contain.
292 3. Algorithm L is rather slow for a hash algorithm, however it's easy.
294 4. Chaining is best in terms of speed; ordered/self-ordering is even
297 5. Rehashing is slow.
299 6. Might want to decide on an algorithm empirically since there are no
300 clear mathematical winners in some cases.
302 7. gprof? Hey, it works!
304 MORE NOTES/IDEAS/BUGS
305 ---------------------
307 The behavior of converting a floating point to an integer when the value of the
308 float is out of range of the integer type is UNDEFINED! See ANSI 6.2.1.3.
310 What should we do for *negative* times in expressions?
312 Sometimes very wide (or very tall) columns can occur in tables. What is a good
313 way to truncate them? It doesn't seem to cause problems for the ascii or
314 postscript drivers, but it's not good in the general case. Should they be
315 split somehow? (One way that wide columns can occur is through user request,
316 for instance through a wide PRINT request--try time-date.stat with a narrow
317 ascii page or with the postscript driver on letter size paper.)
319 NULs in input files break the products we're replacing: although it will input
320 them properly and display them properly as AHEX format, it truncates them in A
321 format. Also, string-manipulation functions such as CONCAT truncate their
322 results after the first NUL. This should simplify the result of PSPP design.
323 Perhaps those ugly a_string, b_string, ..., can all be eliminated.
325 From Moshe Braner <mbraner@nessie.vdh.state.vt.us>: An idea regarding MATCH
326 FILES, again getting BEYOND the state of SPSS: it always bothered me that if I
327 have a large data file and I want to match it to a small lookup table, via
328 MATCH FILES FILE= /TABLE= /BY key, I need to SORT the large file on key, do the
329 match, then (usually) re-sort back into the order I really want it. There is
330 no reason to do this, when the lookup table is small. Even a dumb sequential
331 search through the table, for every case in the big file, is better, in some
332 cases, than the sort. So here's my idea: first look at the /TABLE file, if it
333 is "small enough", read it into memory, and create an index (or hash table,
334 whatever) for it. Then read the /FILE and use the index to match to each case.
335 OTOH, if the /TABLE is too large, then do it the old way, complaining if either
336 file is not sorted on key.
338 ----------------------------------------------------------------------
339 Statistical procedures:
341 For each case we read from the input program:
343 1. Execute permanent transformations. If these drop the case, stop.
344 2. N OF CASES. If we have already written N cases, stop.
345 3. Write case to replacement active file.
346 4. Execute temporary transformations. If these drop the case, stop.
347 5. Post-TEMPORARY N OF CASES. If we have already analyzed N cases, stop.
348 6. FILTER, PROCESS IF. If these drop the case, go to 5.
349 7. Pass case to procedure.
353 LAG records cases in step 3.
355 AGGREGATE: When output goes to an external file, this is just an ordinary
356 procedure. When output goes to the active file, step 3 should be skipped,
357 because AGGREGATE creates its own case sink and writes to it in step 7. Also,
358 TEMPORARY has no effect and we just cancel it. Regardless of direction of
359 output, we should not implement AGGREGATE through a transformation because that
360 will fail to honor FILTER, PROCESS IF, N OF CASES.
362 ADD FILES: Essentially an input program. It silently cancels unclosed LOOPs
363 and DO IFs. If the active file is used for input, then runs EXECUTE (if there
364 are any transformations) and then steals vfm_source and encapsulates it. If
365 the active file is not used for input, then it cancels all the transformations
366 and deletes the original active file.
372 MATCH FILES: Similar to AGGREGATE. This is a procedure. When the active file
373 is used for input, it reads the active file; otherwise, it just cancels all the
374 transformations and deletes the original active file. Step 3 should be
375 skipped, because MATCH FILES creates its own case sink and writes to it in step
376 7. TEMPORARY is not allowed.
384 UPDATE: same as ADD FILES.
387 ----------------------------------------------------------------------
390 * Before TEMPORARY, limits number of cases sent to the sink.
392 * After TEMPORARY, limits number of cases sent to the procedure.
394 * Without TEMPORARY, those are the same cases, so it limits both.
398 * Sample is just a transformation. It has no special properties.
402 * Always selects cases sent to the procedure.
404 * No effect on cases sent to sink.
406 * Before TEMPORARY, selection is permanent. After TEMPORARY,
407 selection stops after a procedure.
411 * Always selects cases sent to the procedure.
413 * No effect on cases sent to sink.
415 * Always stops after a procedure.
419 * Ignored by AGGREGATE. Used when procedures write matrices.
421 * Always applies to the procedure.
423 * Before TEMPORARY, splitting is permanent. After TEMPORARY,
424 splitting stops after a procedure.
428 * TEMPORARY has no effect on AGGREGATE when output goes to the active file.
430 * SORT CASES, ADD FILES, RENAME VARIABLES, CASESTOVARS, VARSTOCASES,
431 COMPUTE with a lag function cannot be used after TEMPORARY.
433 * Cannot be used in DO IF...END IF or LOOP...END LOOP.
435 * FLIP ignores TEMPORARY. All transformations become permanent.
437 * MATCH FILES and UPDATE cannot be used after TEMPORARY if active
438 file is an input source.
440 * RENAME VARIABLES is invalid after TEMPORARY.
442 * WEIGHT, SPLIT FILE, N OF CASES, FILTER, PROCESS IF apply only to
443 the next procedure when used after TEMPORARY.
447 * Always applies to the procedure.
449 * Before TEMPORARY, weighting is permanent. After TEMPORARY,
450 weighting stops after a procedure.
453 -------------------------------------------------------------------------------