[From_CTB] ktable.get(wrong_length_string) gives core dump

[RamRS]

Transferring #14 opened on 13-Sep-2012 (0 comments)

Example using latest khmer from github under 64bit Linux,

$ python
Python 2.7.3 (default, Sep 13 2012, 13:18:09) 
[GCC 4.4.6 20120305 (Red Hat 4.4.6-4)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import khmer
>>> ktable = khmer.new_ktable(6)
>>> ktable.consume("ATGAGAGACACAGGGAGAGACCCAATTAGAGAATTGGACC")
35
>>> ktable.get("ATGAGA")
1
>>> ktable.get("GATGAG")
0
>>> ktable.get("ATAGA")
python: ktable.cc:22: khmer::HashIntoType khmer::_hash(const char*, khmer::WordLength, khmer::HashIntoType&, khmer::HashIntoType&): Assertion `strlen(kmer) >= k' failed.
Aborted (core dumped)

I would expect the output to be zero, or perhaps a ValueError exception about the kmer length being wrong.

[mr-c]

Huzzah: this test case now returns zero.

Unless @ctb or @peterjc have another test case then we can close this.

[RamRS]

I might take a look into this next...

[luizirber]

This looks related to #308

[RamRS]

Yep. Maybe test once again after #308 is merged?

Ram

[luizirber]

@mr-c required test cases for #308 before merging, it will happen sometime this week. I hope.

[RamRS]

Cool, will watch out for it, then. Thank you, @luizirber !

[mr-c]

The KTable class is unused by the current codebase. I purpose that it is removed.

@ctb what do you think?

[ctb]

On Wed, Mar 19, 2014 at 03:11:36PM -0700, Michael R. Crusoe wrote:

The KTable class is unused by the current codebase. I purpose that it is removed.

@ctb what do you think?

Fine by me.

[mr-c]

Note: table.get("wrong-sized-string") still blows up with table = khmer.new_hashbits(...), table = khmer.new_hashtable(...), and table = khmer.new_counting_hash(...)

In [3]: table = khmer.new_counting_hash(6,1e6)

In [4]: table.consume("ATGAGAGACACAGGGAGAGACCCAATTAGAGAATTGGACC")                                                                           
Out[4]: 35

In [5]: table.get("ATGAGA")                                                                                                                 
Out[5]: 1

In [6]: table.get("GATGAG")                                                                                                                
Out[6]: 0

In [7]: table.get("ATAGA")
terminate called after throwing an instance of 'khmer::khmer_exception'
  what():  Generic khmer exception
Aborted

[mr-c]

When this is fixed the following should be done provide a better error message:

diff --git a/lib/kmer_hash.cc b/lib/kmer_hash.cc
index 403ef80..c0ed2c6 100644
--- a/lib/kmer_hash.cc
+++ b/lib/kmer_hash.cc
@@ -27,7 +27,7 @@ HashIntoType _hash(const char * kmer, const WordLength k,
 {
     // sizeof(HashIntoType) * 8 bits / 2 bits/base
     if (!(k <= sizeof(HashIntoType)*4) || !(strlen(kmer) >= k)) {
-        throw khmer_exception();
+        throw khmer_exception("Supplied kmer string doesn't match the underlying k-size.");
     }

     HashIntoType h = 0, r = 0;

[Echelon9]

I've got a fix for this one. PR coming.

[Echelon9]

This issue should now be resolved. Thanks @mr-c for merging the PR fix.

[peterjc]

I originally filed this as #14 on ctb/khmer, but seems that got lost in the GitHub repository move as the new issue 14 is unrelated.

[ctb]

thx @peterjc - do we need to do anything here?