/* dassdr.f -- translated by f2c (version 19980913). You must link the resulting object file with the libraries: -lf2c -lm (in that order) */ #include "f2c.h" /* Table of constant values */ static integer c__4 = 4; static integer c__3 = 3; static integer c__1 = 1; static integer c__256 = 256; static integer c__0 = 0; /* $Procedure DASSDR ( DAS, segregate data records ) */ /* Subroutine */ int dassdr_(integer *handle) { /* Initialized data */ static integer next[3] = { 2,3,1 }; static integer prev[3] = { 3,1,2 }; /* System generated locals */ integer i__1, i__2, i__3; /* Builtin functions */ integer s_rnge(char *, integer, char *, integer); /* Local variables */ integer base; char crec[1024]; doublereal drec[128]; integer free, irec[256], lrec, dest; logical more; integer unit, type__, i__, j, n; extern /* Subroutine */ int chkin_(char *, ftnlen); integer ncomc; extern /* Subroutine */ int maxai_(integer *, integer *, integer *, integer *); char savec[1024]; doublereal saved[128]; integer recno, savei[256]; extern integer sumai_(integer *, integer *); integer ncomr, total, lword, count[4], ltype, start; extern logical failed_(void); extern /* Subroutine */ int dasadi_(integer *, integer *, integer *), cleari_(integer *, integer *); integer drbase; extern /* Subroutine */ int dasioc_(char *, integer *, integer *, char *, ftnlen, ftnlen), dasiod_(char *, integer *, integer *, doublereal *, ftnlen), dasllc_(integer *), dasrdi_(integer *, integer *, integer *, integer *), dashfs_(integer *, integer *, integer *, integer *, integer *, integer *, integer *, integer *, integer *), dasudi_(integer *, integer *, integer *, integer *); integer minadr, maxadr, scrhan, lastla[3]; extern /* Subroutine */ int dassih_(integer *, char *, ftnlen), dashlu_( integer *, integer *), daswbr_(integer *), dasrri_(integer *, integer *, integer *, integer *, integer *); integer offset; extern /* Subroutine */ int dasioi_(char *, integer *, integer *, integer *, ftnlen); integer lastrc[3]; extern /* Subroutine */ int dasops_(integer *), dasufs_(integer *, integer *, integer *, integer *, integer *, integer *, integer *, integer *, integer *), chkout_(char *, ftnlen); integer lastwd[3], nresvc; extern logical return_(void); integer nresvr, savtyp, prvtyp, loc, pos; /* $ Abstract */ /* Segregate the data records in a DAS file into clusters, using */ /* one cluster per data type present in the file. */ /* $ Copyright */ /* Copyright (1995), California Institute of Technology. */ /* U.S. Government sponsorship acknowledged. */ /* $ Required_Reading */ /* DAS */ /* $ Keywords */ /* DAS */ /* FILES */ /* ORDER */ /* SORT */ /* $ Declarations */ /* $ Brief_I/O */ /* Variable I/O Description */ /* -------- --- -------------------------------------------------- */ /* HANDLE I DAS file handle. */ /* $ Detailed_Input */ /* HANDLE is a file handle of a DAS file opened for writing. */ /* $ Detailed_Output */ /* None. See $Particulars for a description of the effect of this */ /* routine. */ /* $ Parameters */ /* None. */ /* $ Exceptions */ /* 1) If the input file handle is invalid, the error will be */ /* diagnosed by routines called by this routine. */ /* 2) If a Fortran read attempted by this routine fails, the */ /* error will be diagnosed by routines called by this routine. */ /* The state of the DAS file undergoing re-ordering will be */ /* indeterminate. */ /* 3) If a Fortran write attempted by this routine fails, the */ /* error will be diagnosed by routines called by this routine. */ /* The state of the DAS file undergoing re-ordering will be */ /* indeterminate. */ /* 4) If any other I/O error occurs during the re-arrangement of */ /* the records in the indicated DAS file, the error will be */ /* diagnosed by routines called by this routine. */ /* $ Files */ /* See the description of the argument HANDLE in $Detailed_Input. */ /* $ Particulars */ /* Normally, there should be no need for routines outside of */ /* SPICELIB to call this routine. */ /* The effect of this routine is to re-arrange the data records */ /* in a DAS file so that the file contains a single cluster for */ /* each data type present in the file: in the general case, there */ /* will be a single cluster of each of the integer, double */ /* precision, and character data types. */ /* The relative order of data records of a given type is not */ /* affected by this re-ordering. After the re-ordering, the DAS */ /* file contains a single directory record that has one descriptor */ /* for each cluster. After that point, the order in the file of the */ /* sets of data records of the various data types will be: */ /* +-------+ */ /* | CHAR | */ /* +-------+ */ /* | DP | */ /* +-------+ */ /* | INT | */ /* +-------+ */ /* Files that contain multiple directory records will have all but */ /* the first directory record moved to the end of the file when the */ /* re-ordering is complete. These records are not visible to the */ /* DAS system and will be overwritten if data is subsequently added */ /* to the DAS file. */ /* The purpose of segregating a DAS file's data records into three */ /* clusters is to make read access more efficient: when a DAS file */ /* contains a single directory with at most three cluster type */ /* descriptors, mapping logical to physical addresses can be done */ /* in constant time. */ /* $ Examples */ /* 1) Segregate data records in a DAS file designated by */ /* HANDLE: */ /* CALL DASSDR ( HANDLE ) */ /* $ Restrictions */ /* None. */ /* $ Literature_References */ /* None. */ /* $ Author_and_Institution */ /* K.R. Gehringer (JPL) */ /* N.J. Bachman (JPL) */ /* W.L. Taber (JPL) */ /* $ Version */ /* - SPICELIB Version 2.0.1 19-DEC-1995 (NJB) */ /* Corrected title of permuted index entry section. */ /* - EKLIB Version 2.0.0, 17-NOV-1993 (KRG) */ /* Added test of FAILED after each DAS call, or sequence of calls, */ /* which returns immediately if FAILED is true. This fixes a bug */ /* where DASOPS signals an error and then DASSDR has a */ /* segmentation fault. */ /* Removed references to specific DAS file open routines in the */ /* $ Detailed_Input section of the header. This was done in order */ /* to minimize documentation changes if the DAS open routines ever */ /* change. */ /* - EKLIB Version 1.2.0, 07-OCT-1993 (NJB) (HAN) (MJS) */ /* Bug fix: call to CLEARD replaced with call to */ /* CLEARI. */ /* - EKLIB Version 1.1.0, 08-JUL-1993 (NJB) (MJS) */ /* Bug fix: extraneous commas removed from argument lists */ /* in calls to DASADI. */ /* - SPICELIB Version 1.0.0, 11-NOV-1992 (NJB) (WLT) */ /* -& */ /* $ Index_Entries */ /* segregate the data records in a DAS file */ /* -& */ /* $ Revisions */ /* - EKLIB Version 2.0.0, 17-NOV-1993 (KRG) */ /* Added test of failed after each DAS call, or sequence of calls, */ /* which returns immediately if FAILED is true. This fixes a bug */ /* where DASOPS signals an error and then DASSDR has a */ /* segmentation fault. */ /* Removed references to specific DAS file open routines in the */ /* $ Detailed_Input section of the header. This was done in order */ /* to minimize documentation changes if the DAS open routines ever */ /* change. */ /* - EKLIB Version 1.2.0, 07-OCT-1993 (NJB) (HAN) (MJS) */ /* Bug fix: call to CLEARD replaced with call to */ /* CLEARI. */ /* - EKLIB Version 1.1.0, 08-JUL-1993 (NJB) */ /* Bug fix: extraneous commas removed from argument lists */ /* in calls to DASADI. This bug had no visible effect on */ /* VAX and Sun systems, but generated a compile error under */ /* Lahey Fortran. */ /* -& */ /* SPICELIB functions */ /* Local parameters */ /* Data type parameters */ /* Directory pointer locations (backward and forward): */ /* Directory address range location base */ /* Location of first type descriptor */ /* Local variables */ /* Saved variables */ /* NEXT and PREV map the DAS data type codes to their */ /* successors and predecessors, respectively. */ /* Initial values */ /* Standard SPICE error handling. */ if (return_()) { return 0; } else { chkin_("DASSDR", (ftnlen)6); } /* Before starting, make sure that this DAS file is open for */ /* writing. */ dassih_(handle, "WRITE", (ftnlen)5); /* Get the logical unit for this file. */ dashlu_(handle, &unit); if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } /* Write out any buffered records that belong to the file. */ daswbr_(handle); if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } /* We're going to re-order the physical records in the DAS file, */ /* starting with the first record after the first directory. */ /* The other directory records are moved to the end of the file */ /* as a result of the re-ordering. */ /* The re-ordering algorithm is based on that used in the REORDx */ /* routines. To use this algorithm, we'll build an order vector */ /* for the records to be ordered; we'll construct this order vector */ /* in a scratch DAS file. First, we'll traverse the directories */ /* to build up a sort of inverse order vector that tells us the */ /* final destination and data type of each data record; from this */ /* inverse vector we can easily build a true order vector. The */ /* cycles of the true order vector can be traversed without */ /* repetitive searching, and with a minimum of assignment of the */ /* contents of data records to temporary variables. */ /* Allocate a scratch DAS file to keep our vectors in. */ dasops_(&scrhan); if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } /* Now build up our `inverse order vector'. This array is an */ /* inverse order vector only in loose sense: it actually consists */ /* of an integer array that contains a sequence of pairs of integers, */ /* the first of which indicates a data type, and the second of which */ /* is an ordinal number. There is one pair for each data record in */ /* the file. The ordinal number gives the ordinal position of the */ /* record described by the number pair, relative to the other records */ /* of the same type. Directory records are considered to have type */ /* `directory', which is represented by the code DIR. */ /* We also must maintain a count of records of each type. */ cleari_(&c__4, count); /* Get the file summary for the DAS file to be segregated. */ dashfs_(handle, &nresvr, &nresvc, &ncomr, &ncomc, &free, lastla, lastrc, lastwd); if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } /* Find the record and word positions LREC and LWORD of the last */ /* descriptor in the file, and also find the type of the descriptor */ /* LTYPE. */ maxai_(lastrc, &c__3, &lrec, &loc); lword = 0; for (i__ = 1; i__ <= 3; ++i__) { if (lastrc[(i__1 = i__ - 1) < 3 && 0 <= i__1 ? i__1 : s_rnge("lastrc", i__1, "dassdr_", (ftnlen)431)] == lrec && lastwd[(i__2 = i__ - 1) < 3 && 0 <= i__2 ? i__2 : s_rnge("lastwd", i__2, "dassd" "r_", (ftnlen)431)] > lword) { lword = lastwd[(i__1 = i__ - 1) < 3 && 0 <= i__1 ? i__1 : s_rnge( "lastwd", i__1, "dassdr_", (ftnlen)434)]; ltype = i__; } } /* The first directory starts after the last comment record. */ recno = nresvr + ncomr + 2; while(recno <= lrec && recno > 0) { /* Read the directory record. */ dasrri_(handle, &recno, &c__1, &c__256, irec); if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } /* Increment the directory count. */ ++count[3]; /* Add the data type (`directory') and count (1) of the current */ /* record to the inverse order vector. */ dasadi_(&scrhan, &c__1, &c__4); dasadi_(&scrhan, &c__1, &count[3]); if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } /* Set up our `finite state machine' that tells us the data */ /* types of the records described by the last read directory. */ type__ = irec[8]; prvtyp = prev[(i__1 = type__ - 1) < 3 && 0 <= i__1 ? i__1 : s_rnge( "prev", i__1, "dassdr_", (ftnlen)478)]; /* Now traverse the directory and update the inverse order */ /* vector based on the descriptors we find. */ more = TRUE_; i__ = 10; while(more) { /* Obtain the count for the current descriptor. */ n = (i__2 = irec[(i__1 = i__ - 1) < 256 && 0 <= i__1 ? i__1 : s_rnge("irec", i__1, "dassdr_", (ftnlen)492)], abs(i__2)); /* Update our inverse order vector to describe the positions */ /* of the N records described by the current descriptor. */ i__1 = n; for (j = 1; j <= i__1; ++j) { dasadi_(&scrhan, &c__1, &type__); i__3 = count[(i__2 = type__ - 1) < 4 && 0 <= i__2 ? i__2 : s_rnge("count", i__2, "dassdr_", (ftnlen)501)] + j; dasadi_(&scrhan, &c__1, &i__3); if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } } /* Adjust the count of records of data type TYPE. */ count[(i__1 = type__ - 1) < 4 && 0 <= i__1 ? i__1 : s_rnge("count" , i__1, "dassdr_", (ftnlen)513)] = count[(i__2 = type__ - 1) < 4 && 0 <= i__2 ? i__2 : s_rnge("count", i__2, "dass" "dr_", (ftnlen)513)] + n; /* Find the next type. */ ++i__; if (i__ > 256 || recno == lrec && i__ > lword) { more = FALSE_; } else { if (irec[(i__1 = i__ - 1) < 256 && 0 <= i__1 ? i__1 : s_rnge( "irec", i__1, "dassdr_", (ftnlen)527)] > 0) { type__ = next[(i__1 = type__ - 1) < 3 && 0 <= i__1 ? i__1 : s_rnge("next", i__1, "dassdr_", (ftnlen)528)]; } else if (irec[(i__1 = i__ - 1) < 256 && 0 <= i__1 ? i__1 : s_rnge("irec", i__1, "dassdr_", (ftnlen)530)] < 0) { type__ = prev[(i__1 = type__ - 1) < 3 && 0 <= i__1 ? i__1 : s_rnge("prev", i__1, "dassdr_", (ftnlen)531)]; } else { more = FALSE_; } } } /* The forward pointer in this directory tells us where the */ /* next directory record is. When there are no more directory */ /* records, this pointer will be zero. */ recno = irec[1]; } /* At this point, the inverse order vector is set up. The array */ /* COUNT contains counts of the number of records of each type we've */ /* seen. Set TOTAL to the total number of records that we've going */ /* to permute. */ total = sumai_(count, &c__4); /* The next step is to build a true order vector. Let BASE be */ /* the base address for the order vector; this address is the */ /* last logical address of the inverse order vector. */ base = total << 1; /* We'll store the actual order vector in locations BASE + 1 */ /* through BASE + TOTAL. In addition, we'll build a parallel array */ /* that contains, for each element of the order vector, the type of */ /* data corresponding to that element. This type vector will */ /* reside in locations BASE + TOTAL + 1 through BASE + 2*TOTAL. */ /* Before setting the values of the order vector and its parallel */ /* type vector, we'll allocate space in the scratch DAS file by */ /* zeroing out the locations we plan to use. After this, locations */ /* BASE+1 through BASE + 2*TOTAL can be written to in random access */ /* fashion using DASUDI. */ i__1 = total << 1; for (i__ = 1; i__ <= i__1; ++i__) { dasadi_(&scrhan, &c__1, &c__0); } if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } /* We note that the way to construct the inverse of a permutation */ /* SIGMA in a single loop is suggested by the relation */ /* -1 */ /* SIGMA ( SIGMA(I) ) = I */ /* We'll use this method. In our case, our order vector plays */ /* the role of */ /* -1 */ /* SIGMA */ /* and the `inverse order vector' plays the role of SIGMA. We'll */ /* exclude the first directory from the order vector, since it's */ /* an exception: we wish to reserve this record. Since the first */ /* element of the order vector (logically) contains the index 1, we */ /* can ignore it. */ i__1 = total; for (i__ = 2; i__ <= i__1; ++i__) { i__2 = (i__ << 1) - 1; i__3 = (i__ << 1) - 1; dasrdi_(&scrhan, &i__2, &i__3, &type__); i__2 = i__ << 1; i__3 = i__ << 1; dasrdi_(&scrhan, &i__2, &i__3, &dest); if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } /* Set DEST to the destination location, measured as an offset */ /* from the last comment record, of the Ith record by adding */ /* on the count of the predecessors of the block of records of */ /* TYPE. */ for (j = 1; j <= 3; ++j) { if (type__ > j) { dest += count[(i__2 = j - 1) < 4 && 0 <= i__2 ? i__2 : s_rnge( "count", i__2, "dassdr_", (ftnlen)628)]; } } /* The destination offset of each record should be incremented to */ /* allow room for the first directory record. However, we don't */ /* need to do this for directory records; they'll already have */ /* this offset accounted for. */ if (type__ != 4) { ++dest; } /* The value of element DEST of the order vector is I. */ /* Write this value to location BASE + DEST. */ i__2 = base + dest; i__3 = base + dest; dasudi_(&scrhan, &i__2, &i__3, &i__); /* We want the ith element of the order vector to give us the */ /* number of the record to move to position i (offset from the */ /* last comment record), but we want the corresponding element */ /* of the type array to give us the type of the record currently */ /* occupying position i. */ i__2 = base + i__ + total; i__3 = base + i__ + total; dasudi_(&scrhan, &i__2, &i__3, &type__); if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } } /* Ok, here's what we've got in the scratch file that's still of */ /* interest: */ /* -- In integer logical addresses BASE + 1 : BASE + TOTAL, */ /* we have an order vector. The Ith element of this */ /* vector indicates the record that should be moved to */ /* location DRBASE + I in the DAS file we're re-ordering, */ /* where DRBASE is the base address of the data records */ /* (the first directory record follows the record having this */ /* index). */ /* -- In integer logical addresses BASE + TOTAL + 1 : BASE + */ /* 2*TOTAL, we have data type indicators for the records to */ /* be re-ordered. The type for the Ith record in the file, */ /* counted from the last comment record, is located in logical */ /* address BASE + TOTAL + I. */ drbase = nresvr + ncomr + 1; /* As we traverse the order vector, we flip the sign of elements */ /* we've accessed, so that we can tell when we encounter an element */ /* of a cycle that we've already traversed. */ /* Traverse the order vector. The variable START indicates the */ /* first element to look at. Ignore the first element; it's a */ /* singleton cycle. */ start = 2; while(start < total) { /* Traverse the current cycle of the order vector. */ /* We `make a hole' in the file by saving the record in position */ /* START, then we traverse the cycle in reverse order, filling in */ /* the hole at the ith position with the record whose number is */ /* the ith element of the order vector. At the end, we deposit */ /* the saved record into the `hole' left behind by the last */ /* record we moved. */ /* We're going to read and write records to and from the DAS file */ /* directly, rather than going through the buffering system. */ /* This will allow us to avoid any untoward interactions between */ /* the buffers for different data types. */ i__1 = base + total + start; i__2 = base + total + start; dasrdi_(&scrhan, &i__1, &i__2, &savtyp); i__1 = base + start; i__2 = base + start; dasrdi_(&scrhan, &i__1, &i__2, &offset); /* Save the record at the location DRBASE + START. */ if (savtyp == 1) { i__1 = drbase + start; dasioc_("READ", &unit, &i__1, savec, (ftnlen)4, (ftnlen)1024); } else if (savtyp == 2) { i__1 = drbase + start; dasiod_("READ", &unit, &i__1, saved, (ftnlen)4); } else { i__1 = drbase + start; dasioi_("READ", &unit, &i__1, savei, (ftnlen)4); } if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } /* Let I be the index of the record that we are going to move */ /* data into next. I is an offset from the last comment record. */ i__ = start; while(offset != start) { /* Mark the order vector element by writing its negative */ /* back to the location it came from. */ i__1 = base + i__; i__2 = base + i__; i__3 = -offset; dasudi_(&scrhan, &i__1, &i__2, &i__3); /* Move the record at location */ /* DRBASE + OFFSET */ /* to location */ /* DRBASE + I */ /* There is no need to do anything about the corresponding */ /* elements of the type vector; we won't need them again. */ /* The read and write operations, as well as the temporary */ /* record required to perform the move, are dependent on the */ /* data type of the record to be moved. */ i__1 = base + total + offset; i__2 = base + total + offset; dasrdi_(&scrhan, &i__1, &i__2, &type__); if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } /* Only pick records up if we're going to put them down in */ /* a location other than their original one. */ if (i__ != offset) { if (type__ == 1) { i__1 = drbase + offset; dasioc_("READ", &unit, &i__1, crec, (ftnlen)4, (ftnlen) 1024); i__1 = drbase + i__; dasioc_("WRITE", &unit, &i__1, crec, (ftnlen)5, (ftnlen) 1024); } else if (type__ == 2) { i__1 = drbase + offset; dasiod_("READ", &unit, &i__1, drec, (ftnlen)4); i__1 = drbase + i__; dasiod_("WRITE", &unit, &i__1, drec, (ftnlen)5); } else { i__1 = drbase + offset; dasioi_("READ", &unit, &i__1, irec, (ftnlen)4); i__1 = drbase + i__; dasioi_("WRITE", &unit, &i__1, irec, (ftnlen)5); } if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } } /* OFFSET is the index of the next order vector element to */ /* look at. */ i__ = offset; i__1 = base + i__; i__2 = base + i__; dasrdi_(&scrhan, &i__1, &i__2, &offset); i__1 = base + i__ + total; i__2 = base + i__ + total; dasrdi_(&scrhan, &i__1, &i__2, &type__); if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } } /* The last value of I is the location in the cycle that element */ /* START followed. Therefore, the saved record corresponding */ /* to index START should be written to this location. */ if (savtyp == 1) { i__1 = drbase + i__; dasioc_("WRITE", &unit, &i__1, savec, (ftnlen)5, (ftnlen)1024); } else if (savtyp == 2) { i__1 = drbase + i__; dasiod_("WRITE", &unit, &i__1, saved, (ftnlen)5); } else { i__1 = drbase + i__; dasioi_("WRITE", &unit, &i__1, savei, (ftnlen)5); } /* Mark the order vector element by writing its negative */ /* back to the location it came from. */ i__1 = base + i__; i__2 = base + i__; i__3 = -start; dasudi_(&scrhan, &i__1, &i__2, &i__3); if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } /* Update START so that it points to the first element of a cycle */ /* of the order vector that has not yet been traversed. This will */ /* be the first positive element of the order vector in a location */ /* indexed higher than the current value of START. Note that */ /* this way of updating START guarantees that we don't have to */ /* backtrack to find an element in the next cycle. */ offset = -1; while(offset < 0 && start < total) { ++start; i__1 = base + start; i__2 = base + start; dasrdi_(&scrhan, &i__1, &i__2, &offset); if (failed_()) { chkout_("DASSDR", (ftnlen)6); return 0; } } /* At this point, START is the index of an element in the order */ /* vector that belongs to a cycle where no routine has gone */ /* before, or else START is the last index in the order vector, */ /* in which case we're done. */ } /* At this point, the records in the DAS are organized as follows: */ /* +----------------------------------+ */ /* | File record | ( 1 ) */ /* +----------------------------------+ */ /* | Reserved records | ( 0 or more ) */ /* | | */ /* +----------------------------------+ */ /* | Comment records | ( 0 or more ) */ /* | | */ /* | | */ /* +----------------------------------+ */ /* | First directory record | ( 1 ) */ /* +----------------------------------+ */ /* | Character data records | ( 0 or more ) */ /* | | */ /* +----------------------------------+ */ /* | Double precision data records | ( 0 or more ) */ /* | | */ /* +----------------------------------+ */ /* | Integer data records | ( 0 or more ) */ /* | | */ /* +----------------------------------+ */ /* | Additional directory records | ( 0 or more ) */ /* | | */ /* +----------------------------------+ */ /* Not all of the indicated components must be present; only the */ /* file record and first directory record will exist in all cases. */ /* The `additional directory records' at the end of the file serve */ /* no purpose; if more data is appended to the file, they will be */ /* overwritten. */ /* The last step in preparing the file is to fill in the first */ /* directory record with the correct information, and to update */ /* the file summary. */ recno = drbase + 1; cleari_(&c__256, irec); /* Set the logical address ranges in the directory record, for each */ /* data type. */ for (type__ = 1; type__ <= 3; ++type__) { maxadr = lastla[(i__1 = type__ - 1) < 3 && 0 <= i__1 ? i__1 : s_rnge( "lastla", i__1, "dassdr_", (ftnlen)937)]; if (maxadr > 0) { minadr = 1; } else { minadr = 0; } irec[(i__1 = type__ << 1) < 256 && 0 <= i__1 ? i__1 : s_rnge("irec", i__1, "dassdr_", (ftnlen)945)] = minadr; irec[(i__1 = (type__ << 1) + 1) < 256 && 0 <= i__1 ? i__1 : s_rnge( "irec", i__1, "dassdr_", (ftnlen)946)] = maxadr; } /* Set the descriptors in the directory. Determine which type */ /* comes first: the order of priority is character, double */ /* precision, integer. */ pos = 9; for (type__ = 1; type__ <= 3; ++type__) { if (lastla[(i__1 = type__ - 1) < 3 && 0 <= i__1 ? i__1 : s_rnge("las" "tla", i__1, "dassdr_", (ftnlen)959)] > 0) { if (pos == 9) { /* This is the first type for which any data is present. */ /* We must enter a type code at position BEGDSC in the */ /* directory, and we must enter a count at position */ /* BEGDSC+1. */ irec[8] = type__; irec[9] = count[(i__1 = type__ - 1) < 4 && 0 <= i__1 ? i__1 : s_rnge("count", i__1, "dassdr_", (ftnlen)969)]; lastrc[(i__1 = type__ - 1) < 3 && 0 <= i__1 ? i__1 : s_rnge( "lastrc", i__1, "dassdr_", (ftnlen)970)] = recno; lastwd[(i__1 = type__ - 1) < 3 && 0 <= i__1 ? i__1 : s_rnge( "lastwd", i__1, "dassdr_", (ftnlen)971)] = 10; pos += 2; prvtyp = type__; } else { /* Place an appropriately signed count at location POS in */ /* the directory. */ if (type__ == next[(i__1 = prvtyp - 1) < 3 && 0 <= i__1 ? i__1 : s_rnge("next", i__1, "dassdr_", (ftnlen)980)]) { irec[(i__1 = pos - 1) < 256 && 0 <= i__1 ? i__1 : s_rnge( "irec", i__1, "dassdr_", (ftnlen)981)] = count[( i__2 = type__ - 1) < 4 && 0 <= i__2 ? i__2 : s_rnge("count", i__2, "dassdr_", (ftnlen)981)]; } else { irec[(i__1 = pos - 1) < 256 && 0 <= i__1 ? i__1 : s_rnge( "irec", i__1, "dassdr_", (ftnlen)983)] = -count[( i__2 = type__ - 1) < 4 && 0 <= i__2 ? i__2 : s_rnge("count", i__2, "dassdr_", (ftnlen)983)]; } lastrc[(i__1 = type__ - 1) < 3 && 0 <= i__1 ? i__1 : s_rnge( "lastrc", i__1, "dassdr_", (ftnlen)986)] = recno; lastwd[(i__1 = type__ - 1) < 3 && 0 <= i__1 ? i__1 : s_rnge( "lastwd", i__1, "dassdr_", (ftnlen)987)] = pos; ++pos; prvtyp = type__; } } } /* Since we've done away with all but the first directory, the first */ /* free record is decremented by 1 less than the directory count. */ free = free - count[3] + 1; /* Write out the new directory record. Don't use the DAS buffered */ /* write mechanism; this could trash the file by dumping buffered */ /* records in the wrong places. */ dasioi_("WRITE", &unit, &recno, irec, (ftnlen)5); /* Write out the updated file summary. */ dasufs_(handle, &nresvr, &nresvc, &ncomr, &ncomc, &free, lastla, lastrc, lastwd); /* Clean up the DAS data buffers: we don't want buffered scratch */ /* file records hanging around there. Then get rid of the scratch */ /* file. */ daswbr_(&scrhan); dasllc_(&scrhan); chkout_("DASSDR", (ftnlen)6); return 0; } /* dassdr_ */