Workspace utilities for developers
Contains routines for summarizing, searching, and performing static analysis on the contents of a workspace.
Summarization routines list the names of objects of particular classes (functions, variables, or tables), or provide a type- dependent synopsis of their properties. Names can be specified either explicitly or by referencing a parent namespace (in which case all objects in all namespaces below it are considered). Object size is approximated and does not include attribute overhead.
The summarization routines are listed below.
| Name and Syntax | Description |
|---|---|
.ws.fns[names] |
Displays the names of functions contained in the specified namespaces, or in all namespaces if the argument is ` |
.ws.vars[names] |
Displays the names of variables (excluding tables) contained in the specified namespaces, or in all namespaces if the argument is ` |
.ws.tbls[names] |
Displays the names of tables contained in the specified namespaces, or in all namespaces if the argument is ` |
.ws.fnsum[names] |
Summarizes functions in the argument, expanding namespaces as appropriate. If the argument is `, all functions in all namespaces are included. The function summary includes the function name, its parameter list, line count, and size |
.ws.varsum[names] |
Summarizes variables (excluding tables) in the argument, expanding namespaces as appropriate. If the argument is `, all variables in all namespaces are included. The variable summary includes the variable name, its type, rank, size, and shape. Type is represented by the Q type character, with uppercase indicating a nonscalar value (e.g. a vector or higher-dimensional array); l and L indicate an enumerated value, Y a dictionary, and - a heterogeneous object. Rank is the number of leading uniform axes (where a scalar is 0) |
.ws.tblsum[names] |
Summarizes tables in the argument, expanding namespaces as appropriate. If the argument is `, all tables in all namespaces are included. The table summary includes the table name, its type, keys, row count, column count, and size. Type is M for an in-memory table, S for a splayed table, or P for a partitioned table |
fnsum, varsum, and tblsum return keyed tables as results so that one can, for example, compare two samples.
q).ws.fns`.
assert assertz buildIrMap cciao chkp chkres chktbl ciao
corroborate cs0 cutByID defer defp dpy eflag enc
expand expp falseAt fmt getNewRz getRa getRd getRz
getvg minToSpan mkvg pen ps pub recutByID
..
The formatting width of the name lists produced by fns, vars, and tbls is controlled by the value of .ws.WTH, which can be reassigned to change.
q).ws.fnsum`.
Function | Params Lines Size
-----------| --------------------------------
assert | [x;y] 1 202
assertz | [x;y] 1 264
buildIrMap | [dc;lts] 14 2433
cciao | [chanID;ts;c] 1 295
chkp | [s;p;q] 3 854
chkres | [r;s] 5 1021
chktbl | [r] 5 1470
ciao | [t;c] 1 525
corroborate| [fn;dc;ii] 43 5802
cs0 | [x] 1 194
cutByID | [chType] 4 831
defer | [t;f;req;x] 5 926
defp | [s;d] 3 514
..
q).ws.varsum`.mdb`g`w`.misc.ZR
Variable | Type Rank Size Shape
-----------------------| --------------------
.mdb.cfg.del | c 0 10 1
.mdb.cfg.destroyMountNS| b 0 10 1
.mdb.cfg.eodTS | p 0 17 1
.mdb.cfg.mountNS | s 0 11 1
.mdb.cfg.ns | s 0 14 1
.mdb.cfg.ops | Y 1 76 ,4
.mdb.cfg.tableOrder | S 1 14 ,0
.mdb.cts | p 0 17 1
.mdb.hdbPurview | j 0 17 1
.mdb.nsday | j 0 17 1
.mdb.purviews | P 1 30 ,2
.mdb.rdbPurview | j 0 17 1
.mdb.updLUT | Y 1 404 ,11
.misc.ZR | C 2 337 ,19
g | C 2 32 2 3
w | L 1 18 ,2
q).ws.tblsum`.
Table | Type Keys Rows Cols Size
-------------| --------------------------------------------
ADU | S :: 0 6 98
Account | S :: 340166 12 31295436
Channel | S :: 1010938 19 68744069
ImportLog | P :: 18 12 2833
Location | S :: 392443 18 36610666
MP | S :: 392487 14 21586725
MPParam | M :: 392443 2 7914267
Organization | M :: 66 15 13435
Reading | P :: 793531105 14 50785990843
Tool | S :: 392471 17 31967213
ToolParam | M :: 392443 2 74489224
ToolType | M :: 26 7 3729
TimeZone | S :: 589 2 11380
User | S :: 38 11 2641
attr | M `id`createTS 392443 7 16482726
Call tree routines provide a way to interrogate and graphically display the functions invoked (directly and/or indirectly) by a specified root function. Code tags can be used to provide additional hints regarding referenced objects.
| Name and Syntax | Description |
|---|---|
.ws.calls[fn] |
Lists the names of functions invoked directly by the specified root function |
.ws.rcalls[fn] |
Lists the names of functions invoked directly or indirectly by the specified root function |
.ws.fntree[names] |
Displays the full calling tree of functions in the argument, expanding namespaces as appropriate |
q).ws.NSX:-1_.ws.NSX / For example purposes, allow workspace routines to operate on own namespace
q)count 0N!.ws.calls`.ws.fntree
`s#`.ws.gfns`.ws.rt`.ws.tr
3
q)count .ws.rcalls`.ws.fntree
16
The calls and rcalls routines can be used in conjunction with the Q code profiler to profile
only a specific calling subtree. For example:
q)\l prof.q
q).prof.prof .ws.rcalls`run
The example below shows the output of computing a call tree of the fntree function itself.
q).ws.fntree`.ws.fntree
┌─ .ws.getn ─── .ws.expns ─┌─ .ws.getn+
┌─ .ws.gfns ─│ └─ .ws.ns
│ │─ .ws.mt
│ └─ .ws.val
│─ .ws.rt
│ ┌─ .ws.ct
─ .ws.fntree ─│ │ ┌─ .ws.cref ─┌─ .ws.cref*
│ │ ┌─ .ws.ctiref ─── .ws.lref ─│ └─ .ws.lref+
│ │─ .ws.ctref ─│ └─ .ws.lref*
│ ┌─ .ws.calls ─│ └─ .ws.qn
│ │ │
│ │ │ ┌─ .ws.cref ─┌─ .ws.cref*
└─ .ws.tr ─│ │─ .ws.lref ─│ └─ .ws.lref+
│ │ └─ .ws.lref*
│ └─ .ws.qn
│─ .ws.rt
└─ .ws.tr*
In the output of fntree, * following a name indicates direct recursion, and + indicates indirect recursion.
fnxref performs an independent static analysis and cross-reference of one or more Q
functions, as well as their embedded lambdas. Code tags can be used to
provide additional hints regarding referenced objects.
For each specified function, each object identifier within the function is classified according to its type, and this is followed by a suspicious reference indicator and details about each reference. Reference information consists of the line number on which the reference occurred and the type of reference.
The identifier types are as follows:
| Identifier Type | Description |
|---|---|
* |
Unclassified |
PR |
Parameter |
LV |
Local variable |
GV |
Global variable |
FN |
Function |
RC |
Recursive function |
LM |
Lambda function |
KW |
Q keyword |
If an identifier's usage involves multiple types (for example, as a result of conflicting
usage in an embedded lambda), the identifier type is followed by +.
Aside from the symbolic identifiers used in the function, fnxref includes : as a
special identifier and cross-references it against return statements, if any.
Identifiers are marked as suspicious through the presence of the ? character after the
object type. This may or may not indicate a problem, but it frequently does. Examples
of conditions that earn the suspicious reference indicator include the following:
- unreferenced or duplicate local
- local/global assignment conflict
- local assignment to a global
- global assignment to a local
- assignment to a keyword
Each reference to each identifier is classified as one of the following:
| Reference Type | Description |
|---|---|
| (blank) | Unclassified |
: |
Assignment, e.g. a:1 or a::1 |
[ |
Index reference, e.g. a[2] or function call |
[: |
Index assignment, e.g. a[2]:x or a[2]::x |
@[ |
At amend, e.g. @[a;0;:;10] |
.[ |
Dot amend, e.g. .[a;(0;2 3);:;10] |
⍫: |
Modified assignment, where ⍫ is one of + - * % & \ | < > = ^ ! ~ , # _ $ ? @ . , e.g. a+:x or a-::x |
For example, a reference consisting of just a line number indicates a value is being sampled; a reference followed by : indicates a value is being assign
A6E0
ed.
Note Assignment vs. execution of lambdas or projections is not distinguished, and these are reported as assignments.
.ed.qed is a line editor implemented in Q. The full source is available at ed. We use .ed.qed as a sample
function to illustrate how the cross-reference output relates to the source. We first define a simple function to prefix function lines with ordinal numbers,
to make the cross-reference relationship clearer.
q)dfn:{-1 (5$"[",'string[til count x],'"]"),'x:"\n"vs last value value x;}
q)dfn`.ed.qed
[0] {
[1] if[0~v:ncsv x;:()];nm:first v;c:v 1; / Extract name (with possible namespace) and context
[2] i:0,1+where"\n"=fn0:nm,":",last v; / Find line breaks
[3] Lns::10000*til count Fn::i _fn0,"\n"; / Scaled line numbers and corresponding lines
[4] Cur::0|-2+count Lns; / Current line (= insert point), before closing }
[5] Mode::0b; / Set insert (vs. edit) mode
[6] Ln::-1; / User-specified line number, if any
[7] if[v 2;fn0:0]; / Kill inceptive defn if new
[8] d:system"c";system"c 1000 2000"; / Set display size
[9] p:system"P";system"P 10"; / Set formatting precision
[10]
[11] dl(); / Display all lines
[12]
[13] while[not$[[2 pr:fmtn seln[];"\\w"~s:read0 0];
[14] [s:"";count nm:def[n;ctx[c;nm;n:name Fn];(1+fn?":")_fn:-1_(,/)Fn]];[if[i:"\\q"~s;nm:""];i]]; / Attempt to define function
[15] r:$[0=count s:ltrim s;Cur; / No change if input empty
[16] [Ln::-1;"["=first s];lcmd s; / Look for edit command
[17] upd pr,s]; / Otherwise, update current line
[18] $[r=-1;-2 "Command error";Mode&::r=Cur::r&-1+count Lns]];
[19]
[20] if[count nm;-1 nm,(" defined";" unchanged")fn0~fn];
[21]
[22] system"c ",.Q.s1 d;system"P ",string p; / Restore settings
[23] }
q).ws.fnxref`.ed.qed
.ed.qed:
.Q.s1 FN 22 n LV 14 14:
: * 1 name FN 14
Cur GV 4: 15 18: ncsv FN 1
Fn GV 3: 14 14 nm LV 1: 2 14: 14 14: 20
Ln GV 6: 16: 20
Lns GV 3: 4 18 not KW 13
Mode GV 5: 18&: p LV 9: 22
c LV 1: 14 pr LV 13: 17
count KW 3 4 14 15 18 20 r LV 15: 18 18 18
ctx FN 14[ read0 KW 13
d LV 8: 22 s LV 13: 14: 14 15: 15 16
def FN 14[ 16 17
dl FN 11 seln FN 13[
first KW 1 16 string KW 22
fmtn FN 13 system KW 8 8 9 9 22 22
fn LV 14 14: 20 til KW 3
fn0 LV 2: 3 7: 20 upd FN 17
i LV 2: 3 14: 14 v LV 1: 1 1 2 7
if KW 1[ 7[ 14[ 20[ where KW 2
last KW 2 while KW 13[
lcmd FN 16 x PR 1
ltrim KW 15
A 1-up or 2-up output format is chosen automatically based on the composition of the output (name and line number lengths), and the value of the
configuration parameter .ws.WTH. The default width is 120.
Comments beginning with /# are annotations known as code tags. These comments can be used
to provide semantic hints to programs that recognize them, as well as to surface execution
details that may not be self-evident from visual or static analysis. The call tree and
cross-reference utilities recognize code tags.
The character immediately following /# identifies the type of code tag. The following
types of code tags are supported:
| Code Tag | Description |
|---|---|
/#+ namelist |
Treat names in namelist as if referenced on line |
/#@ namelist |
Treat contents of names in namelist as if referenced on line |
The names listed may or may not be qualified with a namespace. For example:
value r; /#+ run .rep.run0
If a name is not qualified by a namespace, the context of the referencing function is used.
When referencing the contents of a name using /#@, each name must refer to a defined
global variable. For example:
r:cmd[x] . args; /#@ cmd
This generates references to the items within the global variable cmd, with the precise
effect depending upon the type of the variable:
- If the value is a symbol or a list of symbols, the symbols are referenced.
- If the value is a dictionary, then for each entry:
- If the key is a symbol and the value is a lambda, the key symbol qualified by the dictionary name is referenced.
- If the value is a lambda, the lambda is referenced.
- If the value is a symbol or a list of symbols, the value symbols are referenced.
An ordinary comment may appear to the right of a code tag directive, and multiple code tags can appear on the same line:
r:cmd[x] . args; /#@ cmd / Run command
r:(`$"chk",hl)p; /#+ chkhigh chklow / Perform appropriate limit check
REPTAB[c][c;x;y;.rep`stop`start b]; /#@ REPTAB /#+ .rep.start .rep.stop / Invoke report routine
Searching routines provide a way to search functions within a workspace for one or more substrings. As with the summarization routines, names to search can be specified explicitly or by namepsace.
Searching can be either context-free or context-sensitive.
Context-sensitive searches treat certain constructs as
syntactic elements; these constructs are identifier names,
numeric constants (e.g. 1, -0.5, 1b), symbols (e.g. `price),
and character constants (e.g. "Error"). Context-sensitive
searching allows a match of a syntactic element only if the
entire element is matched.
Hits are reported by displaying the entire line on which a
match occurred, with a caret (^) pointing to the first
character of each match on the line. Results are written to
stdout.
The searching routines are listed below.
| Name and Syntax | Description |
|---|---|
.ws.fnshow[names;strs] |
Searches the specified functions or namespaces for one or more strings |
.ws.seshow[names;strs] |
Searches the specified functions or namespaces for one or more strings, matching only syntactic elements |
On the line below, the following syntactic matches are found:
a+b;"a+b";a_b:1;1_b;-1_-1,b;abc-1;101b;4+1 2 3i-1;1i_b;`bid
^ / "a"
^ ^ ^ ^ / "b"
^ / "a+b"
^ ^ ^ ^ / "1"
^ ^ ^ ^ / "-1"
^ / "1 2 3i"
^ / "\"a+b\""
^ / "bid"
/ "i" (no matches)
a+b;"a+b";a_b:1;1_b;-1_-1,b;abc-1;101b;4+1 2 3i-1;1i_b;`bid
q).ws.fnshow[`.;"misc"]
usec (1 occurrence)
{use`.misc`enl`assert`assertz`use`usec,$[x~(::);();x]}
^
q).ws.fnshow[`.misc.mkwdist`.misc.sqz`.misc.zpad;("i";"b";"abs")]
.misc.mkwdist (2 occurrences)
{[p;x;y] (sum each abs[x-\:y]xexp p)xexp 1f%p}
^^
.misc.sqz (3 occurrences)
{-1_x where i|-1_0b,i:" "<>x," "}
^ ^ ^
.misc.zpad (4 occurrences)
{[w;i] $[count i;"0"^neg[w]$string i;0#""]}
^ ^ ^ ^
q).ws.seshow[`.misc.mkwdist`.misc.sqz`.misc.zpad;("i";"b";"abs")]
.misc.mkwdist (1 occurrence)
{[p;x;y] (sum each abs[x-\:y]xexp p)xexp 1f%p}
^
.misc.sqz (2 occurrences)
{-1_x where i|-1_0b,i:" "<>x," "}
^ ^
.misc.zpad (3 occurrences)
{[w;i] $[count i;"0"^neg[w]$string i;0#""]}
^ ^ ^
Leslie Goldsmith