Hashing of nodes

[aiida-bot]

Originally reported by: Andrea Cepellotti (Bitbucket: acepellotti, GitHub: cepellotti)

---

Implement a node hashing. In this way, it would be possible to immediately recognise whether a node has been already stored in the DB, or if a calculation has been run already, which would return a result immediately.
Note also, this would massively speed up the debugging of workflows (repeating many times the same things): we might immediately understand (create a method for it) whether a wf step has calculations to be executed or not.

---

• Bitbucket: https://bitbucket.org/aiida_team/aiida_core/issue/119

[aiida-bot]

Original comment by Andrea Cepellotti (Bitbucket: acepellotti, GitHub: cepellotti):

---

Moreover, to be sure that results are the same, we should make sure that the code and the parsers are not changed. For the code, we should introduce probably an md5sum, and we should also somehow check whether the parser has changed or not.

[aiida-bot]

Original comment by Andrea Cepellotti (Bitbucket: acepellotti, GitHub: cepellotti):

---

As noted by Andrius:
... I would propose reusing existing nodes instead of creating anew on each request to cope with node duplication. I have implemented a measure to control this issue in verdi data {upf,cif} import {upf,cif} by using get_or_create() classmethod and would suggest moving get_or_create() to the Data class. I am not quite sure whether get_or_create() should return only parentless nodes or all and would like to invite for discussion on this. If I understand it correctly, reusing data nodes with parents might introduce connection between even unrelated calculations, that's why I would at first reuse only parentless nodes.