Base class for a payload metadata provider. More...

#include <MetadataProvider.h>

Inheritance diagram for MetadataProvider:

Classes
class	PayloadMetadataCache
	Simple caching structure to keep the payload information for this and the last exp/run. More...

Public Types
using	PayloadMap = std::unordered_map< std::string, PayloadMetadata >
	Type for the internal list of payloads.

Public Member Functions
	MetadataProvider ()=default
	Default constructible.

	MetadataProvider (const std::set< std::string > &usableTagStates)
	Construct with a known set of usable global tag states.

virtual	~MetadataProvider ()=default
	Default destructible.

bool	setTags (const std::vector< std::string > &tags)
	Set the list of globaltag names to be considered for payloads.

bool	getPayloads (int exp, int run, std::vector< PayloadMetadata > &info)
	Update the information in the vector of metadata instances with the actual values.

std::set< std::string >	getUsableTagStates ()
	Get the valid tag states when checking globaltag status.

void	setUsableTagStates (const std::set< std::string > &states)
	Set the valid tag states for this provider when checking globaltag status.

Protected Member Functions
virtual std::string	getGlobaltagStatus (const std::string &name)=0
	Check the status of a global tag with the given name.

virtual bool	updatePayloads (const std::string &globaltag, int exp, int run)=0
	Update the list of existing payloads from a given globaltag, exp and run combination.

void	addPayload (PayloadMetadata &&payload, const std::string &messagePrefix="")
	Add a payload information to the internal list.

void	printInfoMessage (const std::string &provider)
	Print an INFO message about the used metadata provider.

Protected Attributes
std::vector< std::string >	m_tags
	List of globaltags to consider.

std::unordered_map< std::string, PayloadMetadataCache >	m_cache
	Map of globaltag->known metadata.

PayloadMap *	m_payloads {nullptr}
	Map of known payloads for current conditions.

std::set< std::string >	m_usableTagStates {"TESTING", "VALIDATED", "RUNNING", "PUBLISHED"}
	Set of global tag states to consider valid (except for 'INVALID' which is always considered invalid)

Detailed Description

Base class for a payload metadata provider.

This class is supposed to know what payloads exist and if asked will update a given list of PayloadMetadata instances with the actual values valid for the given exp/run.

It will keep an internal cache of all valid payload metadata so that requesting new metadata should be as cheap as possible.

Definition at line 27 of file MetadataProvider.h.

Member Typedef Documentation

◆ PayloadMap

using PayloadMap = std::unordered_map<std::string, PayloadMetadata>

Type for the internal list of payloads.

Definition at line 30 of file MetadataProvider.h.

Constructor & Destructor Documentation

◆ MetadataProvider()

MetadataProvider ( const std::set< std::string > & usableTagStates )

inlineexplicit

Construct with a known set of usable global tag states.

Definition at line 77 of file MetadataProvider.h.

77: m_payloads{nullptr}, m_usableTagStates(usableTagStates) {}

Belle2::Conditions::MetadataProvider::m_usableTagStates

std::set< std::string > m_usableTagStates

Set of global tag states to consider valid (except for 'INVALID' which is always considered invalid)

Definition: MetadataProvider.h:176

Belle2::Conditions::MetadataProvider::m_payloads

PayloadMap * m_payloads

Map of known payloads for current conditions.

Definition: MetadataProvider.h:174

Member Function Documentation

◆ addPayload()

void addPayload	(	PayloadMetadata &&	payload,
		const std::string &	messagePrefix = `""`
	)

protected

Add a payload information to the internal list.

This should be called by implementations during updatePayloads() for each payload found.

Parameters

payload	payload information filled from the globaltag
messagePrefix	a message prefix to be shown for possible log messages to indicate the correct metadata provider

Definition at line 91 of file MetadataProvider.cc.

  {
    const auto [it, inserted] = m_payloads->emplace(payload.name, payload);
    if (!inserted) {
      auto& existing = it->second;
      if (existing.revision < payload.revision) {
        std::swap(existing, payload);
      }
      B2DEBUG(36, messagePrefix << (messagePrefix.empty() ? "" : ": ") << "Found duplicate payload. Discarding one of them"
              << LogVar("globaltag", existing.globaltag)
              << LogVar("name", existing.name)
              << LogVar("revision", existing.revision)
              << LogVar("checksum", existing.checksum)
              << LogVar("discarded revision", payload.revision));
    } else {
      B2DEBUG(37, messagePrefix << (messagePrefix.empty() ? "" : ": ") << "Found payload"
              << LogVar("globaltag", payload.globaltag)
              << LogVar("name", payload.name)
              << LogVar("revision", payload.revision)
              << LogVar("checksum", payload.checksum));
    }
  }

◆ getGlobaltagStatus()

virtual std::string getGlobaltagStatus ( const std::string & name )

protectedpure virtual

Check the status of a global tag with the given name.

Returns "" if the tag doesn't exist or any other error occured

Implemented in LocalMetadataProvider, CentralMetadataProvider, and NullMetadataProvider.

◆ getPayloads()

bool getPayloads	(	int	exp,
		int	run,
		std::vector< PayloadMetadata > &	info
	)

Update the information in the vector of metadata instances with the actual values.

The input is a list of metadata instances and the metadata provider will try to modify them in place and add all the missing information it can find.

It will try to find iov, revision, checksum, url for each entry which doesn't have a valid revision number (revision=0)
It will ignore any entries with a valid revision number (revision>0)
any payload which cannot be found in any globaltag will raise an error as long as required=true
the modified list can still contain payloads without a valid revision some payloads could not be found.

This function will call updatePayloads() if necessary to update the list of known payloads for a given run but it will try to cache these results so incrementally asking for new payloads as they get requested is perfectly fine.

updatePayloads() will only called for globaltags if we still need to find payload information. If all payloads are found no further queries will be made so adding additional globaltags that are not used at the end is not expensive.

Parameters

	exp	the experiment number
	run	the run number
[in,out]	info	a list of metadata instances which we try to find the missing metadata for.

Returns: true if all required payloads could be found, false if at least one payload which still didn't have a valid revision number and is required could not be found.

Definition at line 36 of file MetadataProvider.cc.

  {
    if (info.empty()) return true;
    for (const auto& tag : m_tags) {
      // Check whether we already know the payloads for this exp,run ...
      bool present{false};
      std::tie(present, m_payloads) = m_cache[tag].get(exp, run);
      // Apparently not, try to get them
      if (!present && !updatePayloads(tag, exp, run)) {
        // Error obtaining payloads means something is wrong (server/network/file problem).
        // In this case we want to fall back to next metadata provider ...
        throw std::runtime_error("Problem updating metadata");
      }
      // And once we got them let's update the info for all requested payloads
      const auto& existing = *m_payloads;
      // and check if we found all payloads we're looking for.
      // Warning: **DO NOT** use `std::all_of` here which looks better but stops
      // iteration on the first missing payload instead of going through all the
      // remaining as well.
      const size_t found = std::count_if(info.begin(), info.end(), [&existing](auto & payload) {
        // already filled by previous gt or has a filename from testing payloads ... so don't do anything
        if (payload.revision > 0 or !payload.filename.empty()) return true;
        // otherwise look for the payload in the list of existing payloads for this run
        if (auto&& it = existing.find(payload.name); it != existing.end()) {
          payload.update(it->second);
          B2DEBUG(35, "Found requested payload metadata"
                  << LogVar("globaltag", payload.globaltag)
                  << LogVar("name", payload.name)
                  << LogVar("revision", payload.revision)
                  << LogVar("checksum", payload.checksum));
          return true;
        }
        // Not found, fine, not a big problem yet, there can be multiple global tags
        return false;
      });
      // If we found all we don't need to check the next global tag at all ...
      if (found == info.size()) break;
    }
    // Check that all payloads could be found (unless they're optional)
    // Again, we don't use std::all_of here because then it stops early and we
    // don't get errors for all missing payloads. But at least it would still be correct.
    const int missing = std::count_if(info.begin(), info.end(), [this, exp, run](const auto & p) {
      if (p.revision == 0 and p.filename.empty() and p.required) {
        B2ERROR("Cannot find payload in any of the configured global tags"
                << LogVar("name", p.name)
                << LogVar("globaltags", boost::algorithm::join(m_tags, ", "))
                << LogVar("experiment", exp) << LogVar("run", run));
        return true;
      }
      return false;
    });
    // Only happy without missing payloads
    return missing == 0;
  }

◆ getUsableTagStates()

std::set< std::string > getUsableTagStates ( )

inline

Get the valid tag states when checking globaltag status.

Definition at line 122 of file MetadataProvider.h.

122{ return m_usableTagStates; }

◆ printInfoMessage()

void printInfoMessage ( const std::string & provider = "" )

protected

Print an INFO message about the used metadata provider.

Parameters

provider the actual metadata provider used for the job

Definition at line 114 of file MetadataProvider.cc.

  {
    B2INFO("Conditions Database: found working metadata provider"
           << LogVar("provider", provider));
  }

◆ setTags()

bool setTags ( const std::vector< std::string > & tags )

Set the list of globaltag names to be considered for payloads.

This should be called before requesting any payloads but after calling setValidTagStates().

Returns: true if all globaltags can be found and are in a valid state

Definition at line 15 of file MetadataProvider.cc.

  {
    m_tags = tags;
    const size_t validTags = std::count_if(m_tags.begin(), m_tags.end(), [this](const auto & name) {
      // dispatch obtaining the state to the implementation
      auto status = getGlobaltagStatus(name);
      // empty status: Unspecified error already dealt with
      if (status.empty()) return false;
      // otherwise check for valid states
      if (m_usableTagStates.count(status) == 0) {
        B2ERROR("The globaltag has a status which is not permitted for use. This is for your own protection"
                << LogVar("globaltag", name)
                << LogVar("status", status)
                << LogVar("allowed states", boost::algorithm::join(m_usableTagStates, ", ")));
        return false;
      }
      return true;
    });
    return validTags == tags.size();
  }

◆ setUsableTagStates()

void setUsableTagStates ( const std::set< std::string > & states )

inline

Set the valid tag states for this provider when checking globaltag status.

Should be called before setTags() if necessary.

Warning: The state "INVALID" will never be accepted and removed from the given set if present

Definition at line 128 of file MetadataProvider.h.

    {
      m_usableTagStates = states;
      m_usableTagStates.erase("INVALID");
    }

◆ updatePayloads()

virtual bool updatePayloads	(	const std::string &	globaltag,
		int	exp,
		int	run
	)

protectedpure virtual

Update the list of existing payloads from a given globaltag, exp and run combination.

This function has to be implemented by specializations. It will be called whenever we need to know the list of payloads available for a given globaltag, exp, run combination. The base class will make sure to call this class only if necessary, so hopefully only once per combination but could be more often if processing jumps between runs.

The implementation should then obtain all metadata valid for this globaltag, exp, run combination and call addPayload() once for each payload to add it to the internal cache.

Returns: true on success, false on failure (e.g. network or file problems)

Implemented in CentralMetadataProvider, LocalMetadataProvider, and NullMetadataProvider.

Member Data Documentation

◆ m_cache

std::unordered_map<std::string, PayloadMetadataCache> m_cache

protected

Map of globaltag->known metadata.

Definition at line 172 of file MetadataProvider.h.

◆ m_payloads

PayloadMap* m_payloads {nullptr}

protected

Map of known payloads for current conditions.

Definition at line 174 of file MetadataProvider.h.

◆ m_tags

std::vector<std::string> m_tags

protected

List of globaltags to consider.

Definition at line 170 of file MetadataProvider.h.

◆ m_usableTagStates

std::set<std::string> m_usableTagStates {"TESTING", "VALIDATED", "RUNNING", "PUBLISHED"}

protected

Set of global tag states to consider valid (except for 'INVALID' which is always considered invalid)

Definition at line 176 of file MetadataProvider.h.

The documentation for this class was generated from the following files:

framework/database/include/MetadataProvider.h
framework/database/src/MetadataProvider.cc

Classes

Public Types

Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

Member Typedef Documentation

◆ PayloadMap

Constructor & Destructor Documentation

◆ MetadataProvider()

Member Function Documentation

◆ addPayload()

◆ getGlobaltagStatus()

◆ getPayloads()

◆ getUsableTagStates()

◆ printInfoMessage()

◆ setTags()

◆ setUsableTagStates()

◆ updatePayloads()

Member Data Documentation

◆ m_cache

◆ m_payloads

◆ m_tags

◆ m_usableTagStates