Beware of type Text!

Dave Matthews matthews at GREENGENES.CIT.CORNELL.EDU
Thu Oct 13 18:09:59 EST 1994


I just figured out a bug in GrainGenes that's plagued me for many months.
Since March I've been unable to reload the database from a full dump.  I
had to load it from my original .ace files, and I had to load them in a 
particular order.  If I didn't the usual error message was 

  !!FATAL ERROR: system error 12 Not enough memory
  malloc failure in messalloc: requesting 29360128 bytes

, though there were others, especially

  !! FATAL ERROR 0: !! Bnew node : block overflow

The results were the same for ace1-10 through ace3.1, under SunOS, Solaris
and Linux.

On further investigation I found that there were certain objects in the
database to which no new data could be added.  No additional links,
directly or by XREF, not even an additional few characters of Text.  Such
objects were found in at least three of my classes, Gene, Gene_Class and
Gene_Set.  Here is one of them.

Gene_Set : "Glu-1"
Reference	 "TAG-60-229"
Reference	 "WGS-83-827"
Reference	 "JCS-3-17"
Reference	 "TAG-13-6833"
Reference	 "NAR-13-8729"
Reference	 "NAR-13-6817"
Reference	 "TAG-67-87"
Reference	 "BCG-24-579"
Reference	 "WGP-84-111"
Reference	 "TAG-75-651"
Gene_class	 "Glutenins"
Gene	 "Glu-D1 (Triticum)"
Gene	 "Glu-A1 (Triticum)"
Gene	 "Glu-B1 (Triticum)"
Gene	 "Glu-E1 (Elytrigia)"
Gene	 "Glu-H1 (Hordeum)"
Gene	 "Glu-H^ch^1 (Hordeum)"
Gene	 "Glu-R1 (Secale)"
Gene	 "Glu-R^m^1 (Secale)"
Gene	 "Glu-U1 (Aegilops)"
Gene	 "Glu-V1 (Dasypyrum)"
Gene	 "Glu-H^t^1 (Elymus)"
Comment	 "The Glu-1 loci, which are compound, encode HMW glutenin \
	subunits."
Comment	 "Each Glu-1 locus in hexaploid wheat contains two genes, the \
	products of which were described as 'x-type' and 'y-type' based on \
	differences in molecular weight and isoelectric point (790).  Other \
	evidence has shown these gene products to differ in electrophoretic \
	fingerprint pattern (798) and cysteine content (730), and the genes \
	themselves to differ in nucleotide sequence (1068, 1034, 255).  \
	Although early evidence suggested up to 6 genes in total at each \
	locus (1069, 255), it appears likely that only a single copy of each \
	gene is present at the 1AL, 1BL,  and 1DL loci (340)."
Comment	 "The symbols for the genes within the Glu-1 loci encoding \
	'x-type' and 'y-type' proteins are Glu-1-1 and Glu-1-2, respectively \
	[rather than Glu-1x and Glu-1y (1068)].  The genes are closely linked \
	but recombination has been observed between Glu-B1-1 and Glu-B1-2 \
	with a frequency of 3 in 3,450 (789).  The gene order, relative to \
	the centromere, has not been ascertained."
Comment	 "A system of naming the Glu-A1-1, Glu-A1-2, Glu-B1-1 and \
	Glu-B1-2 alleles in T. turgidum var. dicoccoides is given in 577."
Numbered_reference	 790 "TAG-60-229"
Numbered_reference	 798 "WGS-83-827"
Numbered_reference	 730 "JCS-3-17"
Numbered_reference	 1068 "TAG-13-6833"
Numbered_reference	 1034 "NAR-13-8729"
Numbered_reference	 255 "NAR-13-6817"
Numbered_reference	 1069 "TAG-67-87"
Numbered_reference	 340 "BCG-24-579"
Numbered_reference	 789 "WGP-84-111"
Numbered_reference	 577 "TAG-75-651"


To make a long story short, the problem turned out to be that those Comment
fields were of type Text.  I changed them to ?Text, and now everything is
working.

A question: Is Text a recommended type for any situations?  As I recall it
used to be required for the arguments of Pick_me_to_call, ?Text didn't work.
Anything else?

- Dave 




More information about the Acedb mailing list