1
0
mirror of https://github.com/golang/go synced 2024-10-01 13:18:33 -06:00
go/cmd/oracle/oracle.el

224 lines
8.2 KiB
EmacsLisp
Raw Normal View History

;;;
;;; Integration of the Go 'oracle' analysis tool into Emacs.
;;;
;;; To install the Go oracle, run:
;;; % export GOROOT=... GOPATH=...
;;; % go get code.google.com/p/go.tools/cmd/oracle
;;; % mv $GOPATH/bin/oracle $GOROOT/bin/
;;;
;;; Load this file into Emacs and set go-oracle-scope to your
;;; configuration. Then, find a file of Go source code, enable
;;; go-oracle-mode, select an expression of interest, and press `C-c C-o d'
;;; (for "describe") or run one of the other go-oracle-xxx commands.
;;;
;;; TODO(adonovan): simplify installation and configuration by making
;;; oracle a subcommand of 'go tool'.
(require 'compile)
(require 'go-mode)
(require 'cl)
(defgroup go-oracle nil
"Options specific to the Go oracle."
:group 'go)
(defcustom go-oracle-command (concat (car (go-root-and-paths)) "/bin/oracle")
"The Go oracle command; the default is $GOROOT/bin/oracle."
:type 'string
:group 'go-oracle)
(defcustom go-oracle-scope ""
"The scope of the analysis. See `go-oracle-set-scope'."
:type 'string
:group 'go-oracle)
(defvar go-oracle--scope-history
nil
"History of values supplied to `go-oracle-set-scope'.")
go.tools/oracle: improvements to command set and performance. Command set: - what: an extremely fast query that parses a single file and returns the AST stack, package name and the set of query modes that apply to the current selection. Intended for GUI tools that need to grey out UI elements. - definition: shows the definition of an identifier. - pointsto: the PTA features of 'describe' have been split out into their own command. - describe: with PTA stripped out, the cost is now bounded by type checking. Performance: - The importer.Config.TypeCheckFuncBodies predicate supports setting the 'IgnoreFuncBodies' typechecker flag on a per-package basis. This means we can load dependencies from source more quickly if we only need exported types. (We avoid gcimport data because it may be absent or stale.) This also means we can run type-based queries on packages that aren't part of the pointer analysis scope. (Yay.) - Modes that require only type analysis of the query package run a "what" query first, and restrict their analysis scope to just that package and its dependencies (sans func bodies), making them much faster. - We call newOracle not oracle.New in Query, so that the 'needs' bitset isn't ignored (oops!). This makes the non-PTA queries faster. Also: - removed vestigial timers junk. - pos.go: existing position utilties split out into own file. Added parsePosFlag utility. - numerous cosmetic tweaks. + very basic tests. To do in follow-ups: - sophisticated editor integration of "what". - better tests. - refactoring of control flow as described in comment. - changes to "implements", "describe" commands. - update design doc + user manual. R=crawshaw, dominik.honnef CC=golang-dev, gri https://golang.org/cl/40630043
2013-12-13 08:04:55 -07:00
;; TODO(adonovan): I'd like to get rid of this separate mode since it
;; makes it harder to use the oracle.
(defvar go-oracle-mode-map
(let ((m (make-sparse-keymap)))
go.tools/oracle: improvements to command set and performance. Command set: - what: an extremely fast query that parses a single file and returns the AST stack, package name and the set of query modes that apply to the current selection. Intended for GUI tools that need to grey out UI elements. - definition: shows the definition of an identifier. - pointsto: the PTA features of 'describe' have been split out into their own command. - describe: with PTA stripped out, the cost is now bounded by type checking. Performance: - The importer.Config.TypeCheckFuncBodies predicate supports setting the 'IgnoreFuncBodies' typechecker flag on a per-package basis. This means we can load dependencies from source more quickly if we only need exported types. (We avoid gcimport data because it may be absent or stale.) This also means we can run type-based queries on packages that aren't part of the pointer analysis scope. (Yay.) - Modes that require only type analysis of the query package run a "what" query first, and restrict their analysis scope to just that package and its dependencies (sans func bodies), making them much faster. - We call newOracle not oracle.New in Query, so that the 'needs' bitset isn't ignored (oops!). This makes the non-PTA queries faster. Also: - removed vestigial timers junk. - pos.go: existing position utilties split out into own file. Added parsePosFlag utility. - numerous cosmetic tweaks. + very basic tests. To do in follow-ups: - sophisticated editor integration of "what". - better tests. - refactoring of control flow as described in comment. - changes to "implements", "describe" commands. - update design doc + user manual. R=crawshaw, dominik.honnef CC=golang-dev, gri https://golang.org/cl/40630043
2013-12-13 08:04:55 -07:00
(define-key m (kbd "C-c C-o t") #'go-oracle-describe) ; t for type
(define-key m (kbd "C-c C-o f") #'go-oracle-freevars)
(define-key m (kbd "C-c C-o g") #'go-oracle-callgraph)
(define-key m (kbd "C-c C-o i") #'go-oracle-implements)
go.tools/oracle: improvements to command set and performance. Command set: - what: an extremely fast query that parses a single file and returns the AST stack, package name and the set of query modes that apply to the current selection. Intended for GUI tools that need to grey out UI elements. - definition: shows the definition of an identifier. - pointsto: the PTA features of 'describe' have been split out into their own command. - describe: with PTA stripped out, the cost is now bounded by type checking. Performance: - The importer.Config.TypeCheckFuncBodies predicate supports setting the 'IgnoreFuncBodies' typechecker flag on a per-package basis. This means we can load dependencies from source more quickly if we only need exported types. (We avoid gcimport data because it may be absent or stale.) This also means we can run type-based queries on packages that aren't part of the pointer analysis scope. (Yay.) - Modes that require only type analysis of the query package run a "what" query first, and restrict their analysis scope to just that package and its dependencies (sans func bodies), making them much faster. - We call newOracle not oracle.New in Query, so that the 'needs' bitset isn't ignored (oops!). This makes the non-PTA queries faster. Also: - removed vestigial timers junk. - pos.go: existing position utilties split out into own file. Added parsePosFlag utility. - numerous cosmetic tweaks. + very basic tests. To do in follow-ups: - sophisticated editor integration of "what". - better tests. - refactoring of control flow as described in comment. - changes to "implements", "describe" commands. - update design doc + user manual. R=crawshaw, dominik.honnef CC=golang-dev, gri https://golang.org/cl/40630043
2013-12-13 08:04:55 -07:00
(define-key m (kbd "C-c C-o c") #'go-oracle-peers) ; c for channel
(define-key m (kbd "C-c C-o r") #'go-oracle-referrers)
go.tools/oracle: improvements to command set and performance. Command set: - what: an extremely fast query that parses a single file and returns the AST stack, package name and the set of query modes that apply to the current selection. Intended for GUI tools that need to grey out UI elements. - definition: shows the definition of an identifier. - pointsto: the PTA features of 'describe' have been split out into their own command. - describe: with PTA stripped out, the cost is now bounded by type checking. Performance: - The importer.Config.TypeCheckFuncBodies predicate supports setting the 'IgnoreFuncBodies' typechecker flag on a per-package basis. This means we can load dependencies from source more quickly if we only need exported types. (We avoid gcimport data because it may be absent or stale.) This also means we can run type-based queries on packages that aren't part of the pointer analysis scope. (Yay.) - Modes that require only type analysis of the query package run a "what" query first, and restrict their analysis scope to just that package and its dependencies (sans func bodies), making them much faster. - We call newOracle not oracle.New in Query, so that the 'needs' bitset isn't ignored (oops!). This makes the non-PTA queries faster. Also: - removed vestigial timers junk. - pos.go: existing position utilties split out into own file. Added parsePosFlag utility. - numerous cosmetic tweaks. + very basic tests. To do in follow-ups: - sophisticated editor integration of "what". - better tests. - refactoring of control flow as described in comment. - changes to "implements", "describe" commands. - update design doc + user manual. R=crawshaw, dominik.honnef CC=golang-dev, gri https://golang.org/cl/40630043
2013-12-13 08:04:55 -07:00
(define-key m (kbd "C-c C-o d") #'go-oracle-definition)
(define-key m (kbd "C-c C-o p") #'go-oracle-pointsto)
(define-key m (kbd "C-c C-o s") #'go-oracle-callstack)
(define-key m (kbd "C-c C-o <") #'go-oracle-callers)
(define-key m (kbd "C-c C-o >") #'go-oracle-callees)
m))
;; TODO(dominikh): Rethink set-scope some. Setting it to a file is
;; painful because it doesn't use find-file, and variables/~ aren't
;; expanded. Setting it to an import path is somewhat painful because
;; it doesn't make use of go-mode's import path completion. One option
;; would be having two different functions, but then we can't
;; automatically call it when no scope has been set. Also it wouldn't
;; easily allow specifying more than one file/package.
(defun go-oracle-set-scope ()
"Set the scope for the Go oracle, prompting the user to edit the
previous scope.
The scope specifies a set of arguments, separated by spaces.
It may be:
1) a set of packages whose main() functions will be analyzed.
2) a list of *.go filenames; they will treated like as a single
package (see #3).
3) a single package whose main() function and/or Test* functions
will be analyzed.
In the common case, this is similar to the argument(s) you would
specify to 'go build'."
(interactive)
(let ((scope (read-from-minibuffer "Go oracle scope: "
go-oracle-scope
nil
nil
'go-oracle--scope-history)))
(if (string-equal "" scope)
(error "You must specify a non-empty scope for the Go oracle"))
(setq go-oracle-scope scope)))
(defun go-oracle--run (mode)
"Run the Go oracle in the specified MODE, passing it the
selected region of the current buffer. Process the output to
replace each file name with a small hyperlink. Display the
result."
(if (not buffer-file-name)
(error "Cannot use oracle on a buffer without a file name"))
;; It's not sufficient to save a modified buffer since if
;; gofmt-before-save is on the before-save-hook, saving will
;; disturb the selected region.
(if (buffer-modified-p)
(error "Please save the buffer before invoking go-oracle"))
(if (string-equal "" go-oracle-scope)
(go-oracle-set-scope))
(let* ((filename (file-truename buffer-file-name))
(posflag (if (use-region-p)
(format "-pos=%s:#%d,#%d"
filename
(1- (go--position-bytes (region-beginning)))
(1- (go--position-bytes (region-end))))
(format "-pos=%s:#%d"
filename
(1- (position-bytes (point))))))
;; This would be simpler if we could just run 'go tool oracle'.
(env-vars (go-root-and-paths))
(goroot-env (concat "GOROOT=" (car env-vars)))
(gopath-env (concat "GOPATH=" (mapconcat #'identity (cdr env-vars) ":"))))
(with-current-buffer (get-buffer-create "*go-oracle*")
(setq buffer-read-only nil)
(erase-buffer)
(insert "Go Oracle\n")
(let ((args (append (list go-oracle-command nil t nil posflag mode)
(split-string go-oracle-scope " " t))))
;; Log the command to *Messages*, for debugging.
(message "Command: %s:" args)
(message nil) ; clears/shrinks minibuffer
(message "Running oracle...")
;; Use dynamic binding to modify/restore the environment
(let ((process-environment (list* goroot-env gopath-env process-environment)))
(apply #'call-process args)))
(insert "\n")
(compilation-mode)
(setq compilation-error-screen-columns nil)
;; Hide the file/line info to save space.
;; Replace each with a little widget.
;; compilation-mode + this loop = slooow.
go.tools/oracle: improvements to command set and performance. Command set: - what: an extremely fast query that parses a single file and returns the AST stack, package name and the set of query modes that apply to the current selection. Intended for GUI tools that need to grey out UI elements. - definition: shows the definition of an identifier. - pointsto: the PTA features of 'describe' have been split out into their own command. - describe: with PTA stripped out, the cost is now bounded by type checking. Performance: - The importer.Config.TypeCheckFuncBodies predicate supports setting the 'IgnoreFuncBodies' typechecker flag on a per-package basis. This means we can load dependencies from source more quickly if we only need exported types. (We avoid gcimport data because it may be absent or stale.) This also means we can run type-based queries on packages that aren't part of the pointer analysis scope. (Yay.) - Modes that require only type analysis of the query package run a "what" query first, and restrict their analysis scope to just that package and its dependencies (sans func bodies), making them much faster. - We call newOracle not oracle.New in Query, so that the 'needs' bitset isn't ignored (oops!). This makes the non-PTA queries faster. Also: - removed vestigial timers junk. - pos.go: existing position utilties split out into own file. Added parsePosFlag utility. - numerous cosmetic tweaks. + very basic tests. To do in follow-ups: - sophisticated editor integration of "what". - better tests. - refactoring of control flow as described in comment. - changes to "implements", "describe" commands. - update design doc + user manual. R=crawshaw, dominik.honnef CC=golang-dev, gri https://golang.org/cl/40630043
2013-12-13 08:04:55 -07:00
;; TODO(adonovan): have oracle give us JSON
;; and we'll do the markup directly.
(let ((buffer-read-only nil)
(p 1))
(while (not (null p))
(let ((np (compilation-next-single-property-change p 'compilation-message)))
;; TODO(adonovan): this can be verbose in the *Messages* buffer.
;; (message "Post-processing link (%d%%)" (/ (* p 100) (point-max)))
(if np
(when (equal (line-number-at-pos p) (line-number-at-pos np))
;; np is (typically) the space following ":"; consume it too.
(put-text-property p np 'display "")
(goto-char np)
(insert " ")))
(setq p np)))
(message nil))
(let ((w (display-buffer (current-buffer))))
(balance-windows)
(shrink-window-if-larger-than-buffer w)
(set-window-point w (point-min))))))
(defun go-oracle-callees ()
"Show possible callees of the function call at the current point."
(interactive)
(go-oracle--run "callees"))
(defun go-oracle-callers ()
"Show the set of callers of the function containing the current point."
(interactive)
(go-oracle--run "callers"))
(defun go-oracle-callgraph ()
"Show the callgraph of the current program."
(interactive)
(go-oracle--run "callgraph"))
(defun go-oracle-callstack ()
"Show an arbitrary path from a root of the call graph to the
function containing the current point."
(interactive)
(go-oracle--run "callstack"))
go.tools/oracle: improvements to command set and performance. Command set: - what: an extremely fast query that parses a single file and returns the AST stack, package name and the set of query modes that apply to the current selection. Intended for GUI tools that need to grey out UI elements. - definition: shows the definition of an identifier. - pointsto: the PTA features of 'describe' have been split out into their own command. - describe: with PTA stripped out, the cost is now bounded by type checking. Performance: - The importer.Config.TypeCheckFuncBodies predicate supports setting the 'IgnoreFuncBodies' typechecker flag on a per-package basis. This means we can load dependencies from source more quickly if we only need exported types. (We avoid gcimport data because it may be absent or stale.) This also means we can run type-based queries on packages that aren't part of the pointer analysis scope. (Yay.) - Modes that require only type analysis of the query package run a "what" query first, and restrict their analysis scope to just that package and its dependencies (sans func bodies), making them much faster. - We call newOracle not oracle.New in Query, so that the 'needs' bitset isn't ignored (oops!). This makes the non-PTA queries faster. Also: - removed vestigial timers junk. - pos.go: existing position utilties split out into own file. Added parsePosFlag utility. - numerous cosmetic tweaks. + very basic tests. To do in follow-ups: - sophisticated editor integration of "what". - better tests. - refactoring of control flow as described in comment. - changes to "implements", "describe" commands. - update design doc + user manual. R=crawshaw, dominik.honnef CC=golang-dev, gri https://golang.org/cl/40630043
2013-12-13 08:04:55 -07:00
(defun go-oracle-definition ()
"Show the definition of the selected identifier."
(interactive)
(go-oracle--run "definition"))
(defun go-oracle-describe ()
go.tools/oracle: improvements to command set and performance. Command set: - what: an extremely fast query that parses a single file and returns the AST stack, package name and the set of query modes that apply to the current selection. Intended for GUI tools that need to grey out UI elements. - definition: shows the definition of an identifier. - pointsto: the PTA features of 'describe' have been split out into their own command. - describe: with PTA stripped out, the cost is now bounded by type checking. Performance: - The importer.Config.TypeCheckFuncBodies predicate supports setting the 'IgnoreFuncBodies' typechecker flag on a per-package basis. This means we can load dependencies from source more quickly if we only need exported types. (We avoid gcimport data because it may be absent or stale.) This also means we can run type-based queries on packages that aren't part of the pointer analysis scope. (Yay.) - Modes that require only type analysis of the query package run a "what" query first, and restrict their analysis scope to just that package and its dependencies (sans func bodies), making them much faster. - We call newOracle not oracle.New in Query, so that the 'needs' bitset isn't ignored (oops!). This makes the non-PTA queries faster. Also: - removed vestigial timers junk. - pos.go: existing position utilties split out into own file. Added parsePosFlag utility. - numerous cosmetic tweaks. + very basic tests. To do in follow-ups: - sophisticated editor integration of "what". - better tests. - refactoring of control flow as described in comment. - changes to "implements", "describe" commands. - update design doc + user manual. R=crawshaw, dominik.honnef CC=golang-dev, gri https://golang.org/cl/40630043
2013-12-13 08:04:55 -07:00
"Describe the selected syntax, its kind, type and methods."
(interactive)
(go-oracle--run "describe"))
go.tools/oracle: improvements to command set and performance. Command set: - what: an extremely fast query that parses a single file and returns the AST stack, package name and the set of query modes that apply to the current selection. Intended for GUI tools that need to grey out UI elements. - definition: shows the definition of an identifier. - pointsto: the PTA features of 'describe' have been split out into their own command. - describe: with PTA stripped out, the cost is now bounded by type checking. Performance: - The importer.Config.TypeCheckFuncBodies predicate supports setting the 'IgnoreFuncBodies' typechecker flag on a per-package basis. This means we can load dependencies from source more quickly if we only need exported types. (We avoid gcimport data because it may be absent or stale.) This also means we can run type-based queries on packages that aren't part of the pointer analysis scope. (Yay.) - Modes that require only type analysis of the query package run a "what" query first, and restrict their analysis scope to just that package and its dependencies (sans func bodies), making them much faster. - We call newOracle not oracle.New in Query, so that the 'needs' bitset isn't ignored (oops!). This makes the non-PTA queries faster. Also: - removed vestigial timers junk. - pos.go: existing position utilties split out into own file. Added parsePosFlag utility. - numerous cosmetic tweaks. + very basic tests. To do in follow-ups: - sophisticated editor integration of "what". - better tests. - refactoring of control flow as described in comment. - changes to "implements", "describe" commands. - update design doc + user manual. R=crawshaw, dominik.honnef CC=golang-dev, gri https://golang.org/cl/40630043
2013-12-13 08:04:55 -07:00
(defun go-oracle-pointsto ()
"Show what the selected expression points to."
(interactive)
(go-oracle--run "pointsto"))
(defun go-oracle-implements ()
"Describe the 'implements' relation for types in the package
containing the current point."
(interactive)
(go-oracle--run "implements"))
(defun go-oracle-freevars ()
"Enumerate the free variables of the current selection."
(interactive)
(go-oracle--run "freevars"))
(defun go-oracle-peers ()
"Enumerate the set of possible corresponding sends/receives for
this channel receive/send operation."
(interactive)
(go-oracle--run "peers"))
(defun go-oracle-referrers ()
"Enumerate all references to the object denoted by the selected
identifier."
(interactive)
(go-oracle--run "referrers"))
;; TODO(dominikh): better docstring
(define-minor-mode go-oracle-mode "Oracle minor mode for go-mode
Keys specific to go-oracle-mode:
\\{go-oracle-mode-map}"
nil " oracle" go-oracle-mode-map)
(provide 'go-oracle)