Changes between Version 3 and Version 4 of LAP_Science
- Timestamp:
- Apr 26, 2011, 3:48:28 PM (15 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
LAP_Science
v3 v4 1 1 = Large Area Processing = 2 2 3 = = Database tables ==3 = Database tables = 4 4 5 The lapRuntable defines and groups the reprocessing of a single projection cell5 The '''lapRun''' table defines and groups the reprocessing of a single projection cell 6 6 7 7 || '''name''' || '''type''' || '''note''' || … … 19 19 || dist_group || string || distribution group to give to all dependents || 20 20 21 The lapExptable defines the pool of exposures, and associates them with the lapRuns that use them21 The '''lapExp''' table defines the pool of exposures, and associates them with the lapRuns that use them 22 22 23 23 || '''name''' || '''type''' || '''note''' || … … 31 31 || data_state || string || || 32 32 33 The lapSequencetable groups a set of lapRuns together into a complete sample of the sky.33 The '''lapSequence''' table groups a set of lapRuns together into a complete sample of the sky. 34 34 35 35 || '''name''' || '''type''' || '''note''' || 36 36 || seq_id || integer || || 37 37 || description || string || || 38 == Processing Steps ==39 38 40 = == Run definition ===39 = Processing Steps = 41 40 42 === Chip-Warp processing === 41 42 43 == Run definition == 44 45 A new lapRun is defined by calling `laptool -definerun -seq_id S -projection_cell PC_NAME -ra RA_CENTER -decl DEC_CENTER -radius R -filter F -label L`. This adds a row to lapRun, and then creates a row in lapExp for all exposures that fall within the area on the sky defined by the -ra/-decl/-radius options. If these exposures have been previously processed (and therefore have a row in lapExp associated with a different lapRun), the chip_id for the new lapExp is copied from that previous entry, provided that the private flag is not marked. 46 47 == Chip-Warp processing == 48 49 Defining the run does not queue chipRuns for those exposures that have not been processed previously. This is done by the `lap_science.pl --chip_mode --lap_id LAP` command. This script calls `laptool -pendingexp -lap_id LAP` which returns a list of all the exposures included in this lapRun. Each exposure is checked, and a chipRun queued if the chip_id entry is blank. 50 51 This stage also defines the pairs for differencing, using the object and comment fields to organize the pairs as is done in nightly science. If a pair is found, the associated pair_id fields point at the chip_id of the companion exposure, and the pairwise flag is set to true. If a pair is not found, pair_id is left as NULL, pairwise is set to false, and private is set to true. This ensures that the difference against the quickstack will not be used by other lapRuns. 52 53 Once all the chips have been queued, the state of the lapRun will be polled using the `lap_science.pl --status --lap_id LAP`. This calls `laptool -exposures -lap_id LAP` which returns the status of each exposure at the chip, camera, fake, warp, diff, and magic stages. In addition to gauging completion, exposure pairs are checked against each other to confirm that no faults have occurred. If one member of a pair has an unrecoverable fault, that entry is marked with data_state = 'drop', and is no longer used. The companion of that unrecoverable exposure is set to be private so it will be diffed against the quickstack. 54 55 Setting the lapExp to private "steals" this processing run away from any other lapRun that may want to use it. If lap_science.pl detects that another lapExp with this exp_id/chip_id set has private = true, then it will requeue a chipRun for that exp_id, and record the new chip_id for itself, and set private = true. 56 57 The script will also check the status of any stacks via the `laptool -stacks -lap_id LAP` which returns the status of all associated stacks for this lapRun. This information is used to set a pair of flags for both the quick and final stacks: the defined flag notes that stacks have been queued, and the have flag notes that the stacks have completed and may be used. 58 59 Using the information about the exposures and stacks for this lapRun, `lap_science.pl` will decide what the next step of the processing will be. After all exposures have a warpRun state of full (or have been dropped due to quality), lap_science.pl checks which step is most appropriate: 43 60 44 61 === Quick stacks === 45 62 63 If any lapExp has private = true, then a quickstack is needed. If defined_quick is false, the quickstack is queued, and the appropriate stackAssociation.sass_id is recorded as lapRun.quick_sass_id. 64 46 65 === Diffs === 66 67 If no lapExp has private = true, and they all have valid pair_id values, then they will all be pairwise diffed. If any lapExp has private = true, and have_quick is true (marking that the quickstack has finished), then the quickstack diff for that exposure will be queued as well. 47 68 48 69 === Final stacks === 49 70 71 Once the diffs have completed, the regular stdscience/distribution tools will be used to process all data through the magic and destreak steps. Once these have completed, the warpRun.magicked value is used to determine if a lapExp is ready for the final stacking stage. If all lapExps have a non-zero magicked value, or have a data_state = 'drop', and defined_final is false, then the final stack is queued, and the appropriate stackAssociation.sass_id is recorded as lapRun.final_sass_id. 72 50 73 === Cleanup === 74 75 Once the final stacks have completed, the lapRun is also complete. If have_final is true, then lapRun.state is set to 'full', and all lapExp.active is set to false. Periodically, `lap_science.pl --cleanup` will poll all lapExp, and if all lapExp with a given chip_id have active = false, then the products of all stages generated from that chip_id will be marked for cleanup (STAGE.label = 'goto_cleaned', STAGE.state = 'goto_cleaned'). 51 76 52 77 == Proposed Changes == … … 54 79 === Update === 55 80 81 Even after cleanup, lapExp still records whether an exposure has been fully processed previously. Therefore, instead of queuing a new chipRun for this exposure, the data products may be updated instead. This will be faster than starting from scratch. Implementing this is not significantly difficult, as it requires only relaxing the state restriction in `laptool -definerun` and having `lap_science.pl` know how to update a lapExp with STAGE.state = 'cleaned'. 82 56 83 === Complete Private diffs === 84 85 If an unpaired lapExp spans multiple projection_cells, then if we make the requirement that the quickstacks associated with all of the projection_cells are complete before diffing, then we do not need to make that lapExp private. The difference can then be done across the entire image, instead of only over the overlapping area. This will reduce the amount of processing required, but will likely increase the mask fraction for that lapExp 86
