parcio/julea

Update documentation

Opened this issue · 0 comments

kiraduwe commented

The documentation needs to be checked/updated

kv client:

  • j_kv_new: update description: "Creates a new key-value pair" - maybe rephrase it so something like "creates pointer/handle to kv"
  • j_kv_new & j_kv_new_index: update description: "\param distribution A distribution." the kv does not have a distribution.
  • j_kv_put: update description: "Creates a key-value pair" - maybe "Puts the key and the value in the kv store"
  • j_kv_put: update description: only 3 of 5 params mentioned
  • j_kv_delete: "* \param item An item." does not have an item parameter
  • j_kv_get: update description: only 2 of 4 params mentioned
  • j_kv_get_callback: update description: "Get a key-value pair." seems not fitting
  • j_kv_get_callback: description: explain parameter "JKVGetFunc func"

kv backend:

  • j_kv_put: document backend behaviour for every backend when an already existing key is put. Overwrite or add?

db client:

  • db client: only for schema_get is a batch_execute necessary. explain why selectors and iterators work without batches and how to find out if the operations succeeded as this is normally stated only in the batch.
  • gboolean j_db_schema_get: should not return a boolean but act similar to j_kv_get etc.
    This returned bool gives the impression it indicates whether the schema could be retrieved/whether it existed, when in fact it indicates whether the scheduling of the operation failed.
  • j_db_entry_get_id: update description: explain how the id can be used
  • change data type of _id (= line number in sql table) in backend from uint32 to uint64. As joins across tables are currently not favored because of missing testing etc. all variable blocks for all files are in one table. So there can be a lot of entries in the table.
  • is there an index on the _id field? if not please add one as this is the main identifier for reading blocks
  • j_db_iterator_get_field: "value points to a new allocated memory region. The caller must free this later using g_free." maybe change for consistency with other clients?
  • docu: how to avoid memory leaks in a while iterator loop

distributed object:

  • j_distribution_set: update description: parameters

round robin:

  • distribution_set: update description: parameters
  • distribution_set: why are there start-index and block-size? Add explanation
  • distribution_free: update description: parameters; is it data or distribution?

Write test:

  • check whether putting existing key actually behaves as expected for every backend type (currently overwriting value)