ms_intro - Man Page

The Mini-SEED library provides a framework for manipulation of SEED data records including the unpacking and packing of data records. Functionality is also included for managing waveform data as continuous traces.  All structures of SEED 2.4 data records are supported with the following exceptions: Blockette 2000 opaque data which has an unknown data structure by definition and Blockette 405 which depends on full SEED (SEED including full ASCII headers) for a full data description.

The primary purpose of the library is to hide the details of Mini-SEED in order to allow rapid development of Mini-SEED reading/writing software.  The framework allows everything from manipulation of Mini-SEED on a record-by-record basis to reading of Mini-SEED into continuous trace segments to packing of large continuous traces using a record template.

Certain common tasks have, through library design, been streamlined, for example: reading Mini-SEED records from a file, adding data from unpacked records to a group of traces or packing a group of continuous traces into Mini-SEED records.

The following data encoding formats are supported for both unpacking and packing: ASCII, INT16, INT32, FLOAT32, FLOAT64, STEIM1 and STEIM2. The INT and FLOAT encodings each have two versions for quantities with a different number of bits of representation.  The STEIM decompression produces 32-bit integers; likewise the compression routines require 32-bit integers as input.  The following data encoding formats are supported for unpacking only: GEOSCOPE (24-bit, 16/3 and 16/4 gain ranged), CDSN, SRO and DWWSSN.

A Mini-SEED record is represented in the library using the data structure given below.  This structure is used for both unpacking and packing of Mini-SEED records.  When unpacking with msr_unpack(3) this structure is populated.  When packing with msr_pack(3) this structure is used as a template for the resulting data records and as a source of samples to be packed.

Blockettes following the fixed section of the header are contained in the blockette chain of BlktLink structures.  Shortcut pointers to commonly used blockettes are maintained for types 100, 1000 and 1001.

Many common header fields which are not easily accessible/usable in the raw header are available directly from the structure.  When this structure is used as a packing template, these common header fields are packed into the appropriate place in the fixed section or blockette.  As examples, the ASCII stream identifiers (network, station, location and channel) are available as NULL terminated strings, the start time is available as a high precision epoch time (see ms_time(3)) and the sample rate is available as a double precision floating point value.

The MSRecord data structure:

typedef struct MSRecord_s {
  char           *record;            /* Mini-SEED record */
  int32_t         reclen;            /* Length of Mini-SEED record */
  
  /* Pointers to SEED data record structures */
  struct fsdh_s      *fsdh;          /* Fixed Section of Data Header */
  struct BlktLink    *blkts;         /* Root of blockette chain */
  struct blkt_100_s  *Blkt100;       /* Blockette 100, if present */
  struct blkt_1000_s *Blkt1000;      /* Blockette 1000, if present */
  struct blkt_1001_s *Blkt1001;      /* Blockette 1001, if present */
  
  /* Common header fields in accessible form */
  int32_t         sequence_number;   /* SEED record sequence number */
  char            dataquality;       /* Data quality indicator */
  char            network[11];       /* Network designation */
  char            station[11];       /* Station designation */
  char            location[11];      /* Location designation */
  char            channel[11];       /* Channel designation */
  hptime_t        starttime;         /* Record start time */
  double          samprate;          /* Nominal sample rate (Hz) */
  int64_t         samplecnt;         /* Number of samples in record */
  int8_t          encoding;          /* Data encoding format */
  int8_t          byteorder;         /* Byte order of record */
  
  /* Data sample fields */
  void           *datasamples;       /* Data samples */
  int64_t         numsamples;        /* Number of data samples */
  char            sampletype;        /* Sample type code: a, i, f, d */

  /* Stream oriented state information */
  StreamState    *ststate;           /* Stream processing state information */
}
MSRecord;

The library includes two facilities to manage collections of continuous trace segments, each represented by their top most structure: MSTraceGroup and MSTraceList.  The MSTraceList facility is a next generation version of the MSTraceGroup facility.  Whereas the MSTraceGroup facility uses a single linked list of time segments the MSTraceList facility is slightly more complex with two levels of linked lists and common access pointers.  The advantages are that the MSTraceList structure is faster to populate, especially when there are many segments (gappy data), and the list is always maintained in a sorted order.

MSTraceList data structures allow the grouping of MSTraceID structures which are themselves the root of MSTraceSeg structures, see libmseed.h as a reference to these structures.

MSTraceGroup data structures allow the grouping of MSTrace structures. While a MSTrace structure is normally used to hold trace information and associated data samples it can also be used without data samples as a means to keep trace of data coverage without actual samples.

Numerous routines are provided for basic management of MSTrace structures, including the creation of new MSTrace structures, adding data from Mini-SEED data structures to MSTrace structures, printing trace information, etc.

The MSTraceGroup data structure acts as a very simple place to begin a chain of MSTrace structures and keep track of the number of traces.

The MSTrace and MSTraceGroup data structures:

typedef struct MSTrace_s {
  char            network[11];     /* Network designation */
  char            station[11];     /* Station designation */
  char            location[11];    /* Location designation */
  char            channel[11];     /* Channel designation */
  char            dataquality;     /* Data quality indicator */
  char            type;            /* MSTrace type code */
  hptime_t        starttime;       /* Time of first sample */
  hptime_t        endtime;         /* Time of last sample */
  double          samprate;        /* Nominal sample rate (Hz) */
  int64_t         samplecnt;       /* Num. in trace coverage */
  void           *datasamples;     /* Data samples */
  int64_t         numsamples;      /* Num. samples in datasamples */
  char            sampletype;      /* Sample type code: a, i, f, d */
  void           *prvtptr          /* Private pointer for general use */
  struct MSTrace_s *next;          /* Pointer to next trace */
}
MSTrace;

typedef struct MSTraceGroup_s {
  int32_t           numtraces;     /* Number of MSTraces in trace chain */
  struct MSTrace_s *traces;        /* Root of the trace chain */
}
MSTraceGroup;

All of the log and diagnostic messages emitted by the library functions use the same interface.  The output from this interface can be controlled.  This is useful when the library will be embedded in a larger system with a custom logging facility.  See the man page for more details.

  ms_log() : the central logging facility.  Behavior is controlled by
        the settings specified with ms_loginit().

  ms_loginit() : set the functions and prefixes used for log,
        diagnostic and error messages.

The default destination for log messages is standard output (stdout), while all diagnostic (including error) messages go to standard error (stderr).  Most of the internal messages emmited by the library are considered diagnostic and will, by default, go to standard error.

The default prefix for log and diagnostic messages is nothing. The default prefix for diagnostic error messages is "Error: ".

There are reentrant versions of these functions that operate directly on a logging parameter MSLogParam struct.  These are intended for use in threaded programs or where a complex logging scheme is desired. See the man pages for more details.

Waveform data samples are managed by libmseed in a couple of different formats depending on how they are unpacked or will be packed.  An array of samples is completely represented by an array of sample values, the number of samples and a sample type.  The number of samples is always the actual number of sample values, not the number of bytes needed for storing the values.  Samples can be either ASCII, 32-bit integer, 32-bit floats or 64-bit double precision floats.

Sample types are identified by a single ASCII type character:

"a" - ASCII (8 bits)
"i" - integer (32 bits)
"f" - float (32 bits)
"d" - double (64 bits)

The size of each sample type in bytes is returned by the get_samplesize(3) lookup routine.

The SEED 2.4 standard allows data only SEED (Mini-SEED) to be either in big (most significant byte first) or little (least significant byte first) endian byte order.  Unfortunately it is not well defined what little endian Mini-SEED really means.  While libmseed supports all four combinations of big and little endian header and data the surest way to avoid compatibility problems is to always create big endian Mini-SEED records (header and data).

Reading MiniSEED - how libmseed determines the byte order of a record:

The byte order of a record header is determined by checking if the record start year is a sane value (e.g. between 1920 and 2020).  The byte order of (compressed) data samples is determined by the byte order flag in the Blockette 1000, if a Blockette 1000 is not present the byte order is assumed to be the same as the header.  To force the byte order determination of either the header or data section of a record the following environment variables can be set:

UNPACK_HEADER_BYTEORDER
UNPACK_DATA_BYTEORDER

These variables should be set to either 0 (little endian) or 1 (big endian).  A programmatic equivalent of setting these environment variables is provided via the following macros:

MS_UNPACKHEADERBYTEORDER(X)
MS_UNPACKDATABYTEORDER(X)

Writing MiniSEED - in what byte order libmseed creates records:

Normally the byte order of MiniSEED created by libmseed is controlled via a flag in the API.  This byte order flag determines the ordering for both the header and data sections of a record.  To force the byte order of either the header or data section of a record the following environment variables can be set:

PACK_HEADER_BYTEORDER
PACK_DATA_BYTEORDER

These variables should be set to either 0 (little endian) or 1 (big endian).  A programmatic equivalent of setting these environment variables is provided via the following macros:

MS_PACKHEADERBYTEORDER(X)
MS_PACKDATABYTEORDER(X)

Note that some interpretations of the SEED 2.4 format imply that so-called little endian MiniSEED means that the record header is little endian but that the data section is big endian (as the only defined data encodings must be based on the SEED DDL which, in turn, must be defined in terms of big endian).  Libmseed will not create MiniSEED of this flavor by default but can be configured to do so by setting the environment variables described above appropriately.

Example programs using libmseed are provided in the 'examples' directory of the source code distribution.

One of the most common tasks is to read a file of Mini-SEED records and either perform some action based on the header values or apply some process to the data samples.  This task is greatly simplified by using the library functions ms_readmsr(3) and ms_readtraces(3).  The ms_readmsr(3) routine will open a specified file and return MSRecord structures for each Mini-SEED record it reads from the file.  The ms_readtraces(3) routine will do the same except add all the data read to a MSTraceGroup, this is ideal for quickly reading data for processing.  Both of these routines are able to automatically detect record length.

If your application is not designed to read Mini-SEED from files the library also provides functions to detect and parse Mini-SEED records in memory buffers.  For more information see ms_detect(3) and msr_parse(3).

Skeleton code for reading a file with ms_readmsr(3):

main() {
  MSRecord *msr = NULL;
  int retcode;

  while ( (retcode = ms_readmsr (&msr, filename, 0, NULL, NULL, 1, 0, verbose)) == MS_NOERROR )
    {
       /* Do something with the record here, e.g. print */
       msr_print (msr, verbose);
    }

  if ( retcode != MS_ENDOFFILE )
    ms_log (2, "Cannot read %s: %s0, filename, ms_errorstr(retcode));

  /* Cleanup memory and close file */
  ms_readmsr (&msr, NULL, 0, NULL, NULL, 0, 0);
}

For reading two files with ms_readtraces(3):

main() {
  MSTraceGroup *mstg = NULL;
  int retcode;

  retcode = ms_readtraces (&mstg, filename, 0, -1.0, -1.0, 0, 1, 0, verbose);

  if ( retcode != MS_ENDOFFILE )
    ms_log (2, "Cannot read %s: %s0, filename, ms_errorstr(retcode));

  retcode = ms_readtraces (&mstg, filename2, 0, -1.0, -1.0, 0, 1, 0, verbose);

  if ( retcode != MS_ENDOFFILE )
    ms_log (2, "Cannot read %s: %s0, filename2, ms_errorstr(retcode));

  if ( ! mstg )
    {
      fprintf (stderr, "Error reading file\n");
      return -1;
    }

  /* Do something with the traces here, e.g. print */
  mst_printtracelist (mstg, 0, verbose, 0);

  mst_freegroup (&mstg);
}

Another common task is to create (pack) Mini-SEED records. The library supports packing of Mini-SEED either from MSRecord structures, MSTrace structures or MSTraceGroup collections using, respectively, msr_pack(3), mst_pack(3) or mst_packgroup(3).  In each case the appropriate data structure and parameters are provided to the routine along with a function pointer to a routine that will be called each time a record is complete and should be disposed of.

When packing Mini-SEED records the concept of a record header template is used, the template is always in the form of a MSRecord structure. This allows the calling program to dictate the contents, with a few exceptions, of the header in the final data records.

Skeleton code for creating (packing) Mini-SEED records with mst_pack(3):

static void record_handler (char *record, int reclen, void *srcname) {
  if ( fwrite(record, reclen, 1, outfile) != 1 )
    {
      ms_log (2, "Error writing %s to output file0, (char *)srcname);
    }
}

main() {
  int64_t psamples;
  int precords;
  MSTrace *mst;
  char srcname[50];

  mst = mst_init (NULL);

  /* Populate MSTrace values */
  strcpy (mst->network, "XX");
  strcpy (mst->station, "TEST");
  strcpy (mst->channel, "BHE");
  mst->starttime = ms_seedtimestr2hptime ("2004,350,00:00:00.000000");
  mst->samprate = 40.0;

  /* The datasamples pointer and numsamples counter will be adjusted by
     the packing routine, the datasamples array must be dynamic memory
     allocated by the malloc() family of routines. */
  mst->datasamples = dataptr; /* pointer to 32-bit integer data samples */  
  mst->numsamples = 1234;
  mst->sampletype = 'i';      /* declare type to be 32-bit integers */

  mst_srcname (mst, srcname, 0);

  /* Pack 4096 byte, big-endian records, using Steim-2 compression */
  precords = mst_pack (mst, &record_handler, srcname, 4096, DE_STEIM2,
                       1, &psamples, 1, verbose, NULL);

  ms_log (0, "Packed %"PRId64" samples into %d records0,
             psamples, precords);

  /* Disconnect datasamples pointer, otherwise mst_free() will free it */
  mst->datasamples = NULL;

  mst_free (&mst);
}

msr_unpack(3), ms_time(3) and msr_pack(3)

Chad Trabant
IRIS Data Management Center

msr_addblockette(3), msr_duplicate(3), ms_readmsr(3), msr_host_latency(3), msr_init(3), msr_normalize_header(3), msr_pack(3), msr_print(3), msr_samprate(3), msr_starttime(3), msr_unpack(3), mst_addmsr(3), mst_convertsamples(3), mst_findmatch(3), mst_groupsort(3), mst_init(3), mstl_printtracelist(3), mst_pack(3), mst_printtracelist(3), ms_writemseed(3).