next previous
Up: VALD-2: Progress of the Base


Subsections

3 Software improvements

  VALD has been in operation for more than four years. Its success was largely determined by the choice of the data base architecture. Individual line lists are pipelined through a merging and filtering software which can be configured according to the type and parameters of a request.

The upgrade of the VALD software was driven by our intention to handle additional information requested by users, to accomodate larger data sets and to reduce the response time of the data base.

3.1 Internal data format and data flow in VALD

  Currently, each VALD line list is stored on magnetic disk as one of several compressed files. Each file consists of binary records (one record per transition). When a new line list is to be included to VALD, it is first converted to the internal VALD format (see Sect. 2) and then compressed. In preparation for the future inclusion of molecular lines (in VALD-3) two changes to the internal record format have been made already now. First, the species are identified internally with a number rather than by the atomic number and ionization stage. This unique correspondence between a number and a species is defined in the VALD internal table of species. Second, each record has been extended with two new fields that contain the quality parameter and additional reference (see Sect.2). Both of these fields are extracted by the SHOW LINE extraction option of VALD which is used to obtain all the data on a particular transition from all lists included in the data base (see Paperi).

The data compression is based on the Lempel-Ziv-Welch algorithm (Welch 1984) which was modified to achieve the highest compression ratio (files are compressed to less than 46% of the original size) and to preserve direct access to the data. A high compression ratio is achieved by taking advantage of the fixed record length of the internal data format. To speed up the response time we created a special descriptor file that allows starting decompression of a given list within less than 1000 spectral lines from the requested wavelength rather than from the start of a file.

Any data extraction in VALD is performed using a pipeline of the corresponding filters or "VALD extraction tools''. Presently, two tools can initiate a pipeline: PRESELECT and SHOW LINE (the PRESELECT tool is used to extract the best data for each transition in a certain wavelength range, see Paperi). Both of them are calling special routines to access data base files (open, close and read/decompress functions) which helps to make VALD independent of computer systems, because the content of external files is interpreted as a stream of bits while the components of the pipeline exchange ASCII messages. The success of this strategy was confirmed by porting VALD between DEC and Hewlett-Packard workstations, as well as running VALD over the network in a mixed environment, when different pipeline components are running on different computers. The extraction and merging of data from different lists is directed (as before) by the VALD configuration file (cf. Paperi). The format of this file has been extended to accomodate to the possibility to choose term designations from a specific list. Moreover, each client who uses the data base through remote access facilities (VALD EMS and WWW) has now a possibility to keep a personal copy of the configuration file (see Sect.3.3).

3.2 Mirror site support

  Creation of mirror sites is a logical way to distribute the work load and reduce VALD response time. It is particularly important for massive extraction which is performed, e.g. for model atmosphere calculations. Two VALD mirror sites became operational in 1998: Uppsala Astronomical Observatory (This email address is being protected from spambots. You need JavaScript enabled to view it. ) and Astrophysics Data Facility at NASA Goddard Space Flight Center (This email address is being protected from spambots. You need JavaScript enabled to view it. ). In order to achieve coherence of all VALD servers we developed an additional layer of the software that runs automatically on top of the VALD email service and ensures the synchronization of all sites. Currently, synchronization is performed twice a day which means that within less than a day after registration a new client will have EMS access to all VALD sites.

3.3 Changes to the user interface and extracted data format

  As described in Paperi the VALD-EMS (email service) remains the main user interface, but now it offers several additional options. A VALD-EMS request still consists of two parts: the type of request (SHOW LINE, EXTRACT ALL, EXTRACT ELEMENT and EXTRACT STELLAR) and the parameters of extraction. VALD-2 offers the possibility to specify additional options for a better control of the data extraction. The following options can be specified with all request types except for SHOW LINE, where only the PERSONAL CONFIGURATION option is allowed:

SHORT FORMAT 		 default extraction format; 
LONG FORMAT 		 extract term description for each
		 transition;
PERSONAL 		 
  CONFIGURATION 		 use a personal VALD configuration file
		 (see below) instead of the standard
		 VALD configuration;
HAVE RAD 		 select lines with known radiative
		 damping constant;
HAVE STARK 		 select lines with known Stark damping
		 constant;
HAVE WAALS 		 select lines with known vanderWaals
		 damping constant;
HAVE LANDE 		 select lines with known Land'e factor;
HAVE TERM 		 select lines for which VALD contains
		 term designations.

Multiple options can be combined in arbitrary sequence. For example, the following request will extract all neutral iron lines in a given spectral range for which the vanderWaals damping constant is known and provide term description for upper and lower levels:

begin request
extract element
long format
have waals
6170,6190
Fe 1
end request

The resulting extraction procedure will create a list similar to (some of the columns "...'' have been omitted):

                           ... Damping parameters
Elm Ion  WL(A)     log($gf$) ... Rad.   Stark  Waals
'Fe 1', 6170.5040, -0.654, ... 8.243,-5.588,-7.590,
         '(4F)4p y3D(4F)4d ...   1   1   1   1   1'
'Fe 1', 6171.0060, -1.788, ... 8.377,-4.676,-7.749,
         '(4F)4p y3D(4F)4d ...   1   1   1   1   1'
'Fe 1', 6172.2130, -3.916, ... 8.275,-5.181,-7.808,
         '(4F)4p z5Gs6D)4d ...   1   1   1   1   1'
'Fe 1', 6173.0050, -6.670, ... 4.057,-6.272,-7.878,
         '(4F)4s a5F(5D)sp ...   1   1   1   1   1'
'Fe 1', 6173.0280, -5.957, ... 7.799,-6.335,-7.833,
         's2 b3D    (3P)sp ...   1   1   1   1   1'
'Fe 1', 6173.3410, -3.081, ... 8.223,-6.194,-7.815,
         '(4P)4s a5P(4F)4p ...   1   1   1   1   1'
'Fe 1', 6173.6420, -3.413, ... 8.237,-6.069,-7.631,
         '(4F)4p z5Gs4D)5s ...   1   1   1   1   1'
  ...
 References:
  1. FeI NMT Whaling and Bard & Kock

On top of the VALD-EMS, a World-Wide-Web interface was developed. This interface allows interactive creation of a request to VALD-EMS. It notifies the user of inconsitencies or omissions of input data before converting them to an ordinary VALD-EMS request. Just as for VALD-EMS, access to the WWW interface is limited to registered clients only. The interface will resolve the hostname of the connecting computer and attempt to match the combination of the provided username and resolved hostname against the list of email addresses in the client register. If this is successfull, access is granted to the WWW interface. In this way, the interface also determines the client's email address to which VALD-EMS will send the result of the request. It is therefore important that the client provides user@host addresses to the client register at which the client can receive email.

Another addition to the user interface is the personal configuration file which was primarily dictated by the quick evolution of the content of our data base. With the addition of new line data of higher accuracy it is often impossible to reproduce the same selection with the standard VALD configuration file that controls the merging of different line lists according to the ranking system. One can still get the same selection with the old configuration file, but until VALD-2 a user had no control over the content of the configuration file when using the EMS interface to access the data base. Now each client can choose between the standard configuration, created by VALD experts, or a personal configuration file that is modified only by the owner. A personal configuration file is first created as a copy of the standard configuration in response to the VALD-EMS request that contains the " personal configuration'' option or via the WWW interface. The WWW interface also provides a convenient editing tool. The VALD client can select a subset of the line lists that will be searched for extraction requests, set the ranking of different parameters that determines the merging procedure etc. (see Paperi). The latest copies of the personal configuration files are automatically distributed to all VALD servers as part of the synchronization procedure.

Two changes have been introduced to the format of extracted data: case sensitivity for names of species and the new referencing system. The first one is part of preparation for VALD-3 which will include molecular line lists of astrophysical interest. We intend to keep conventional notations for the names of molecules, therefore case sensitivity is crucial to distinguish e.g. between CO (Cobalt) and CO (carbon oxide). The second change will simplify acknowledgement of laboratory spectroscopy work as was suggested by several of the VALD contributors and users. VALD-2 now also compiles a list of references at the end of every extraction where the source of each item in each transition is given in the line reference field. For example, a response to a particular request may contain:

                                 Damping parameters
Elm Ion  WL(A)    Excit(eV) log($gf$)  ....  Waals
Lande factor   References
'Fe 1', 4804.5190,  3.5730, -2.590,  .... -7.824,
 1.170, '   1   1   1   ....  2   2  '
 ...
 References:
  1. FeI NMT Whaling and Bard & Kock
  2. GFIRON obs. energy level: Fe

where reference numbers ' 1 1 1 1 2 2 2 ' correspond to wavelength, excitation potential of the lower level, oscillator strength, 3 damping constants and Landé factor(s). One more reference number for term designation will appear in the long output format. It will be useful in case one develops an automatic term interpretation system since the designations are (unfortunately) not standardized across the lists.

In some cases a single line list may contain a compilation from different sources (like the reference 1 in the example above). For such lists the VALD-2 SHOW LINE request returns a label pointing to the original publication for each datum of each transition. The description of the labels can be found in the reference section of the VALD-EMS electronic document. This document is regularly sent to VALD clients to inform them about new extraction capabilities and/or changes of the user interface. It can also be obtained from the WWW sites of VALD. A label description which compiles all the lists within VALD-2 (as of December 1998) can also be found in Ryabchikova et al. (1999b).


next previous
Up: VALD-2: Progress of the Base

Copyright The European Southern Observatory (ESO)