bnmcgn/gadgets

Ben McGunigle's utility collection

gadgets's Issues

Some systems failed to build for Quicklisp dist

Building with SBCL 2.0.5 / ASDF 3.3.1 for quicklisp dist creation.

Trying to build commit id 58c9505

gadgets fails to build with the following error:

; caught ERROR:
;   READ error during COMPILE-FILE: Package SPLIT-SEQUENCE does not exist. Line: 245, Column: 48, File-Position: 9209 Stream: #<SB-INT:FORM-TRACKING-STREAM for "file /home/quicklisp/quicklisp-controller/dist/build-cache/gadgets/c8761e662f4bd76adb2ec21773bf71a7dcd9c03b/gadgets-20210402-git/gadgets.lisp" {1007C6E3E3}>
...
Unhandled UIOP/LISP-BUILD:COMPILE-FILE-ERROR in thread #<SB-THREAD:THREAD "main thread" RUNNING {1000A10083}>: COMPILE-FILE-ERROR while compiling #<CL-SOURCE-FILE "gadgets" "gadgets">

test-gadgets fails to build because of a failure in gadgets.

Full log here

Issues in the tests in t/test-gadgets.lisp

1. The test with do-hash-table effectively assumes that do-hash-table scans the table in a specific order, but do-hash-table uses alexandria:hash-table-keys, which does not return the keys in a specific order, so it sometimes fails and sometimes doesn't.

2. The test for use-package-with-shadowing assuming that the call to use-package-with-shadowing happens before reading the symbols in the following forms. That is true wen you load the source file, but not if you compile the file and load it, in which case the reading happens at compile time, and the call later at load time. It is better to do the reading at runtime e.g.

                    (is 2 (symbol-value (read-from-string "test1::var1")))

3. Not really a problem, but make it easier to deal with errors: the macro prove-test:is is expecting the test form to be the first argument, and the expected result to be the second argument, so the for above should really be:

                    (is  (symbol-value (read-from-string "test1::var1"))  2)

When the form is first argument, if it causes an error it catches it and goes on the following tests. When it is second, it enters the debugger.

Implementation-indpendent string-unless-number

gadgets:string-unless-number currently uses conditionalization, which limits where it works. I think the following code would do what is required using onlu common-lisp:

(defun string-unless-number (x)
  "Return the input as a string unless it can be recognized as an integer."
  (if (numberp x)
      x
    (if (stringp x)
        (or (ignore-errors (parse-integer x)) x)
      (error "STRING-UNLESS-NUMBER: argument is not a number or a string : ~s" x))))