Discussion List Archives

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ddlm-group] DDLm aliases (subject changed). .. .. .. .. .

Thanks Herbert for your comments.  However, you have not addressed my two main concerns:

1. In a CIF dictionary written under DDLm, is there
a) a single alias_definition_set loop in its own save frame that contain all the aliased datanames, i.e., the properties of the individual definition_sets are not attrubutes of a dataname,
or b) is there an alias_definition_set loop in each data-item save frame, i.e., the alias_defintion_set properties are explicit data-item attributes?

If the answer is a) how does one denormalize since there is a single alias_definition_set loop, but very many parent alias loops in the dictionary.  Does the information in the alias_definition_set loop get distributed as attributes of the different data items?  However, as your draft now reads, there would be little point in denormalizing because the alias loop contains exactly the same information as the alias_definition_set loop, so denormalizing would add no new information.  Collecting all the aliases in one place might make programming a little easier, but should we be designing dictionaries in which the same information appears in two different places, just organized in a different way?  Can't the computer do this for us at run time?  What happens if the two loops disagree with each other?

If the answer to question 1 is b), the alias and alias_definition_set loops contain identical information (or at least it should be identical).

2. Do you envisage that:
a) we will have a number of flavours of mmCIF dictionaries depending on how the definition_sets are defined?  That is, do we have standard_mmCIF (no defnition_sets), mmCIF_a_le_Bernstein, mmCIF_a_le_Westbrook etc.?

or b) the information about the definition_sets will be held in a subdictionary that will be imported at run time to create a virtual dictionary?  This would avoid the problems of different flavours, but it is not clear to me that the import features as currently defined in DDLm would be able to support this.  How do you see such a merge taking place?

I apologize if these are trivial questions, but at some point you will have to instruct the likes of me how the system works, so you might as well do it now :)


Herbert J. Bernstein wrote:
OK, now to the specifics of this item.  I will interpolate my remarks
into David's message with "***" flags. -- Herbert

 Herbert J. Bernstein, Professor of Computer Science
   Dowling College, Kramer Science Center, KSC 121
        Idle Hour Blvd, Oakdale, NY, 11769


On Mon, 31 Jan 2011, David Brown wrote:

Not having done any programming for many years and not being familiar with
the current jargon, I find keeping up with the rapid-fire discussion on
_aliases requires some effort on my part and the discussion has usually
move on before I havc a chance to comment.

However, being incommunicado on weekends, I took Herbert's draft home where
I could sit down without distraction and see what eactly what was
proposed.  The errors and ambiguities in the draft did not help me get my
heasd around the proposal, but I made some progress, and here are my
comments.  First the general comments, then particular comments interleaved
in the draft.

1. _identifier_set:  What does this set identify?  The _id seems to flag
arbitrary groups aliases.  'alias_set' would be a better name for this
category, or even just 'set'.  What's in a name?  It sets certain switches
in the brain, and if these are misset by the dataname, it may take a lot of
work to get them reset.  Not conducive to instant communication. 

*** Yes, for maximal generality, a set is an arbitrary grouping.  We
have been through many alternative names for these groupings, and so far
"set" is the only one to which there has not (yet) been strong objection
on grounds of confusion with other, existing aggregational terminology.
The only (weak) point in favor of "_identifier_set" rather than
"_alias_set" is that the primary save frame name/tag/identifier
can appear in the list.  I would be just as happy with "set", but
then I assume sombody will have been deprived of some future use
of some other set aggregation scheme, so to avoid appearing to
hog a very useful name, it probably should be qualitied in some
way.  Does "tag_set" appeal to people?

2. Why do we need both _alias and _alias_identifier_set categories?  They
have indentical information (if the datanames (Syd's word) or tags
(Herbert's word) are any indication).  I suppose (though this is no where
spelled out) that _identifier_set would have its own save frame in the
dictionary and would not be an attfibute of a datanmae.  If this is a
correct interpretation it would provide a place whare all the alias
datanames in the dictionary could be listed within a single loop).  This
seems redundant, but I cannot speak from programming experience.  If
alias_identifier_set does not appesr in its own save_ frame, how does it
differ from 'alias'?

*** We do not need both an _alias and and _alias_identifier_set category,
and more than we need both an _atom_site and an _atom_site_aniso_U
category (well perhaps a both more).  It is just an organizational
convience to all both the flatter, very denormalized DDL1 style
of presentation as well as the more normalize DDL2 style of
presentation.  It does have the side advantage of allowing
one to pull our a separate list of tags by set.

3. If my supposition in 2. is correct, we would appear to have a problem. 
We will now have a variety of flavours of CIF dictionaries, each expressing
a particular programmer's preferences for grouping the aliases.  This will
make no difference to the CIFs themselves as these groupings are irrelevant
once a CIF has been written, but if, for example, I am given a program
written by Herbert and different program written by either of the Johns, I
might need a different mmCIF dictionary for each of these two programs,
dictionaries that differ only in the way the aliases are gouped.  When I
load my CIF it cannot give instructions on which dictionary to call up
because is will have no knowledge of the idiosyncrasies of the program I
have chosen to use.  A possible solution would be to use the _import
feature to create a virtual dictioanry at run time.  Thus the
_identifier_set information would be held in a local dictionary that would
be imported into the authorised CIF dictioanry at run time.  However, there
are limitations on what can be imported.  The _identifier_set category
could be imported but it would be impossible to import the
_identifier_set_id into the alias loop by this mechanism.  Since I am not
sure that I understand how Herbert intends to use this feature, I do not
feel competent to suggest a way in which this problem could be handled. 
Having programs that used different dialects of CIF dictionaries does not
seem to be in line with the traditional development of CIF, particularly
for a feature that is unlikely to be much used, even if they do not affect
the CIFs themselves.  We should think carefully about the implications of
this move.

*** The main reason I am trying to bring this to COMCIFS is that we
very much seem headed in the direction of multiple, conflicting
interpretations of our standards, and dictionaries that will
not fit together.  Without this, we seem headed towards having,
for example, at least three distinct and possibly conflicting
core dictionaries in simultaneous use (the current core, the one
in mmCIF and the new DDLm core), and at least three and possibly
four mmCUF dictionaries (the official IUCr DDL2 mmCIF, the PDB's
DDL2 pdbx mmCIF, and 2 more DDLm versions of each of those).

4. My detailed comments follow the feature they comment on below:

       _definition.id             '_definition.xref_code'
       _definition.update           2011-01-26
       _definition.class            Attribute
        Code identifying the equivalent definition in the dictionary
        referenced by the DICTIONARY_XREF attributes.

        Use of _definition.xref_code is deprecated in favor of
        use of _alias.xref_code
       _name.category_id            definition
       _name.object_id              xref_code
       _type.purpose                Identify
       _type.container              Single
       _type.contents               Code

This item is deprecated.  It should be deleted.   The  alias and xref
iitems have never been tested and it is clear from the current DDLm that
thay are placeholders that are awaiting development.  If there are any
programs that make use of the items they can only have been written by
members of this group.  The current xref defintiions inadequate and
unworkable. .The is no excuse for leaving this item in if we don't need it.

*** See my general remarks on communicating with the unknown users
of what has been posted since 2007.


       _definition.id               alias
       _definition.scope            Category
       _definition.class            List
       _definition.update           2011-01-26
        The attributes used to specify the aliased names of definitions.
        Every tag has an implicit alias to itself with a null
        _alias.xref_code to allow use of the primary tag in
        the ALIAS_IDENTIFIER_SET category.

        The use of _alias.identifier_set_id in the key of
        this catgeory is provide a placeholder for the
        to conform the key of the parent ALIAS category
        to the key of the child ALIAS_IDENTIFIER_SET
        for automatic joins.  It is not intended that
        _alias.identifier_set_id should be used in the
        ALIAS category when no join is being done.

This last paragraph would be easier to undestand if all the words were
present and the sentences grammatical..  In any case it should appear under
_alias,identifierr_set, not here.  If I am right in thinking alias is an
attribute and alias_identifier_set is not,  how does one join a
non-attribute to an attribute?

*** This stray item has to do with being able to have both the
denormalized flat presentation and the normalized presentation
via a join.  If we stay with just the normalized presentation,
it is not necessary.  I apologize for my typos.  I should have
said "is to provide" instead of "is provide".  Discussing the
key of a category in the category definition seems appropriate
because the key is defined here (see just below).

       _category.parent_id          ddl_attr
       _category_key.primitive      ['_alias.definition_id',

       _definition.id             '_alias.definition_id'
       _definition.class            Attribute
       _definition.update           2006-11-16
        Identifier tag of an aliased definition.
       _name.category_id            alias
       _name.object_id              definition_id
       _type.purpose                Key
       _type.container              Single
       _type.contents               Tag

       _definition.id             '_alias.deprecated'
       _definition.class            Attribute
       _definition.update           2006-11-16
        Specifies whether use of the alias is deprecated
       _name.category_id            alias
       _name.object_id              definition_id

.object_id should be the second part of the _definition.id, i.e.,
'deprecated'.  This needs correcting in many places.

*** thanks for spotting that

       _type.purpose                STATE
       _type.container              Single
       _type.contents               YesorNo
       _enumeration.default         No

       _definition.id             '_alias.dictionary_uri'
       _definition.update           2011-01-26
       _definition.class            Attribute
        Dictionary URI in which the aliased definition belongs.
        _alias.dictionary_uri is deprecated in favor if
       _name.category_id            alias
       _name.object_id              dictionary_uri
       _type.purpose                Identify
       _type.container              Single
       _type.contents               Uri

This item should be moved to the _dictionary_xref category.  The xref.id 
is sufficient link.  Again, the fact that it may appear in the draft
dictionaries should not prevent it being deleted in DDLm since the draft
dictionaries are just drafts and will be chaniged once we have sorted out
how to do the aliases.

*** I donlt disagree on the objective.  I just disagree on whether
the ghost of tags past needs to remain here to haunt its old abode.

       _definition.id   '_alias.identifier_set.id'
       _definition.class  Attribute
       _definition.update 2011-01-26
       A code identifying an identifier_set of related tags.
       This linked item is provided in the ALIAS category to
       ensure that the key of the ALIAS category is
       conformed to the key of the ALIAS_IDENTIFIER_SET
       category.  The alias has not been joined with
       ALIAS_IDENTIFIER_SET, _alias.identifier_set_id
       it is not intended that  _alias.identifier_set_id
       in the ALIAS category.

I cannot make much sense of the last sentence.  Perhaps there are missing

*** Thanks.  I'll try to clean up the words.

       This is a pointer to _identifier_set.id
       _name.category_id alias
       _name.object_id   code

See above.  'code' is not the proper .object_id
*** thanks.  you are right

       _name.linked_item_id         '_identifier_set.id'
       _type.purpose     Key
       _type.container   Single
       _type.contents    Code
       _enumeration.default  .

       _definition.id             '_alias.xref_code'
       _definition.update           2011-01-26
       _definition.class            Attribute
        Code identifying the dictionary containing the primary
        definition of the dictionary as given in the
        DICTIONARY_XREF category.

       _name.category_id            definition
       _name.object_id              xref_code
       _name.linked_item_id         '_dictionary_xref.code'
       _type.purpose                Key
       _type.container              Single
       _type.contents               Code


     _definition.id      identifier_set
     _definition.scope   Category
     _definition.class   List
     _definition.update  2011-01-27
      Data items used to describe the identifier_set identifiers
      used in this dictionary.  Data items in this category
      are NOT used directly as attributes of individual data items.
      See linked item _alias_identifier_set.identifier_set_id
      for such uses.

      _category.parent_id ddl_attr
      _category_key.generic  '_identifier_set.id'


       _definition.id   '_identifier_set.id'
       _definition.class  Attribute
       _definition.update 2011-01-27
       A code identifying an identifier_set of related tags.
       The coverage of an identfier_set may conform precisely
       to the set of tags in a particular dictionary,
       or to tags drawn from multiple dictionaries or
       to a subset of tags from a single dictionary.

       The same tag may belong to multiple identifier
       sets, and a given tag may not belong to any
       identifier set, in which case the only associated
       identifier set is a null value.

Presumably the second line should read 'need not belong to any'.  The
wording above is ambiguous.
The last line should read 'identifier set id' and its default value should
be given explicitly.  What is nul?
Possible nul values are 'nul', '0', ' ', '?', '.', etc.

*** I was thinking of "." and "?"

       To help ensure that dictionaries can be merged,
       each code should either begin with an IUCr-registered
       prefix or, if not prefixed, have been approved
       by COMCIFS.  The special prefix 'local_' may be
       use for purely internal purposes of an organization.

I assume these are not datanames that appear in the dictionaries but a list
of COMCIFS  enumerations, some of which might appear in a non-exclusive
enumeration list.  What happens if someone chooses 'joeblow' as an id?

If COMCIFS approved joeblow, then so be it, but the idea is that
COMCIFS would approved and control the use of set identifiers such
as say DDL1, DDL2, DDLm, core, and mmCIF, but that the PDB, which
has registered the pdbx_ prefix, would control the use of set
identifiers such as pdbx_mmCIF or pdbx_EM, and that somebody who
is doing something purely local, migh just use local_joeblow without
consulting anybody.

       _name.category_id identifier_set
       _name.object_id   code
       _type.purpose     Key
       _type.container   Single
       _type.contents    Code


       _definition.id   '_identifier_set.description'
       _definition.class  Attribute
       _definition.update 2011-01-27
       A description of the identifier_set
       _name.category_id identifier_set
       _name.object_id   code
       _type.purpose     Describe
       _type.container   Single
       _type.contents    Text



      _definition.id      alias_identifier_set
      _definition.scope   Category
      _definition.class   List
      _definition.update  2011-01-27
       The attributes used to specify the identifier_set of
       tags to which a given tag belong.

       A given tag may belong to multiple identifier_sets
       and may be cited against multiple dictionaries.

       Note that _alias_identifier_set.identifier_set_id is a
       component of the key of ALIAS_IDENTIFIER_SET.  If the
       denormalized join presentation is used to bring the object
       ids of this child category up into the parent
       ALIAS category, then _alias.identifier_set_id will
       we used as an implicit addition to the key of
       the denormalized ALIAS category.

       Until DDLm can be formally revised to automatically
       handle the necessary promotion of child catgeory keys
       in denormalized joins, a place-holder
       _alias.identifier_set_id has been defined in the
       ALIAS catgeory.

      _category.parent_id  alias
      _category.parent_join  Yes
      _category_key.primitive  ['_alias_identifier_set.identifier_set_id',

       _definition.id   '_alias_identifier_set.definition_id'
       _definition.class  Attribute
       _definition.update 2011-01-27
       Together with _alias_identifier_set.xref_code, identifies
       an alias belonging to an identifier_set.  An alias may
       belong to any number of identifier_sets, including zero.

       _name.category_id alias_identifier_set
       _name.object_id   definition_id
       _name.linked_item_id  '_alias.definition_id'
       _type.purpose     Key
       _type.container   Single
       _type.contents    Tag

       _definition.id   '_alias_identifier_set.identifier_set_id'
       _definition.class  Attribute
       _definition.update 2011-01-27
       Identifies an identifier_set to which the alias
       identified by _alias_identifier_set.definition_id
       and _alias_identifier_set.xref_code ) belongs.

       A pointer to _identifier_set.id
       _name.category_id alias_identifier_set
       _name.object_id   code
       _name.linked_item_id  '_identifier_set.id'
       _type.purpose     Key
       _type.container   Single
       _type.contents    Code


       _definition.id   '_alias_identifier_set.xref_code'
       _definition.class  Attribute
       _definition.update 2011-01-21
       A code identifying the actual dictionary,
       virtual dictionary or other logical grouping
       to which the identifier tag belongs.

What is this identifier tag - '.definiton_id' or '.definiton_set_id'?

*** They are just pointers, see the primary definitions at
the _name.linked_item_id linked tags.  But for clarity, I
will repeate them locally in the next pass.
       _name.category_id alias_identifier_set
       _name.object_id   code
       _name.linked_item_id  '_dictionary_xref.code'
       _type.purpose     Key
       _type.container   Single
       _type.contents    Code

We also need to refined the _dictionary_xref category.  '.uri; should be
added, '.format' should be better derfined or deleted.  Perhaps '.version'
should also be added.  Definining the dictionaries is just as important as
definiting the definition_sets

*** I'll try to propose something on the next pass, probably on Wednesday
during the next storm.


_______________________________________________ ddlm-group mailing list ddlm-group@iucr.org http://scripts.iucr.org/mailman/listinfo/ddlm-group

fn:I.David Brown
org:McMaster University;Brockhouse Institute for Materials Research
adr:;;King St. W;Hamilton;Ontario;L8S 4M1;Canada
title:Professor Emeritus
tel;work:+905 525 9140 x 24710
tel;fax:+905 521 2773

ddlm-group mailing list

Reply to: [list | sender only]
International Union of Crystallography

Scientific Union Member of the International Council for Science (admitted 1947). Member of CODATA, the ICSU Committee on Data. Member of ICSTI, the International Council for Scientific and Technical Information. Partner with UNESCO, the United Nations Educational, Scientific and Cultural Organization in the International Year of Crystallography 2014.

ICSU Scientific Freedom Policy

The IUCr observes the basic policy of non-discrimination and affirms the right and freedom of scientists to associate in international scientific activity without regard to such factors as ethnic origin, religion, citizenship, language, political stance, gender, sex or age, in accordance with the Statutes of the International Council for Science.